Effect of Daily Fiber Intake Among Cirrhotic Patients With and Without Portosystemic Shunts

Abstract

Background

A diet rich in fiber, especially soluble fiber, causes cholestatic liver damage and fibrosis in animal models with intestinal dysbiosis, high serum bile acid concentrations, and congenital portosystemic shunts (PSs), but no data on patients with cirrhosis (CIRs) are available.

Objectives

To investigate whether dietary fiber consumption was associated with clinical outcomes of CIRs and whether their effect differed according to the presence of PSs.

Methods

Daily soluble and insoluble fiber intake was extrapolated from 3-d food diaries in 25 patients with chronic hepatitis (CH) and 80 CIRs outpatient liver transplant candidates abstinent from alcohol and nonviremic for ≥6 mo. In CIRs, the presence of PSs was verified by computed tomography, and the model for end-stage liver disease (MELD) score was calculated at enrollment and after 6 mo.

Results

PSs were present in 48 (60%) CIRs. The MELD score after 6 mo, compared with enrollment, had improved in 19 and 10 CIRs with and without PSs, respectively. By adjusting for confounders in logistic regression models we found that improvement in MELD over time was inversely associated with insoluble fiber consumption expressed in milligrams per kilogram (mg/kg) body weight in CIRs without PSs [odds ratio (OR): 0.968; 95% confidence interval (CI): 0.939, 0.997; P = 0.005] but with soluble fiber consumption in CIRs with PSs [OR: 0.946; 95% CI: 0.912, 0.982; P = 0.001]. In CIRs with PSs, soluble fiber consumption was inversely associated with normal serum alkaline phosphatase values at enrollment [OR: 0.964; 95% CI: 0.963, 0.993; P = 0.010]. CHs with normal serum alanine transaminase consumed significantly more soluble fiber (p=0.015) than those with abnormal alanine transaminase.

Conclusions

The clinical impact of dietary fiber changes from beneficial to harmful as the stage of chronic liver disease progresses. In particular, in the advanced cirrhosis stage with PSs, soluble fiber intake appears to significantly influence disease progression and should be kept low.

Introduction

Dietary consumption of fiber is considered beneficial for health in general and with regard to various pathologies such as type 2 diabetes, cardiovascular diseases, and cancer [1]. These beneficial effects would be mediated by the ability of dietary fibers to increase the production of metabolites of the intestinal microbiota, which are absorbed and positively influence the host’s metabolism, such as, for example, short-chain fatty acids and secondary bile acids [2]. Both insoluble and soluble fibers regulate the relative abundance of the intestinal microbiota, in turn favoring the production of the aforementioned metabolites [1,3,4]. Furthermore, soluble fibers such as inulin and pectin themselves constitute the substrate from which the intestinal microbiota produces short-chain fatty acids through their fermentation [2].

Even with regards to the liver, few studies conducted on the general population have reported beneficial effects of fiber consumption, reducing the risk of developing and dying from chronic liver disease [5,6]. Only 2 studies involved patients with cirrhosis (CIR), finding a beneficial effect of fiber consumption in patients with relatively compensated disease or in patients with minimal hepatic encephalopathy (HE) [7,8].

However, recent data have called into question whether a high-fiber diet is always beneficial for the liver. In fact, a large study demonstrated no association between fiber consumption and mortality from nonalcoholic fatty liver disease [9]. Furthermore, a meta-analysis found no association between soluble fiber consumption and indices of hepatocyte damage or cholestasis [10]. Indeed, in some healthy subjects, high consumption of soluble fibers can cause intestinal dysbiosis and an acute increase in serum alanine transaminase (ALT) [11]. Furthermore, data on animals demonstrate that a diet enriched in soluble fibers and, to a lesser extent, insoluble fibers can cause chronic damage to hepatocytes and cholangiocytes, liver fibrosis, and hepatocellular carcinoma [[12], [14]]. These liver pathologies develop when a diet rich in fiber is associated with intestinal dysbiosis and high blood concentration of bile acids and are exacerbated when the latter is particularly high, as in the presence of spontaneous portosystemic shunts (PSs) [[12], [14]].

In humans, PSs can be rarely congenital or, more frequently, spontaneously acquired due to portal hypertension in patients with CIR [15,16]. Spontaneous PSs in patients with cirrhosis are associated with increased concentrations of bile acids in the systemic circulation, more severe liver damage, a greater degree of decompensation, and reduced survival [[16], [17], [18], [19]]. To our knowledge, because no studies are available that have considered the impact of dietary fiber content on the clinical outcomes of patients with CIR, also according to the presence or absence of PSs, we studied this aspect in a cohort of outpatient patients with CIR who were candidates for liver transplantation (LT). As clinical outcomes, we considered the values of serum ALT and alkaline phosphatase (ALP) at enrollment and the subsequent temporal changes in the model for end-stage liver disease (MELD) score [20,21].

Methods

Study population

We prospectively enrolled 111 consecutive patients with CIR at their first outpatient visit at the LT outpatient service of the gastroenterology division of La Sapienza University of Rome, Italy. The inclusion criteria were age >18 y and cirrhosis diagnosis made when ≥2 of the following conditions were present: 1) history of cirrhosis complications: HE, variceal gastrointestinal bleeding, and ascites; 2) blood test consistent with cirrhosis: hyperbilirubinemia, hypoalbuminemia, prolonged international normalized ratio, and low platelet count; 3) signs of advanced chronic liver disease and/or portal hypertension at diagnostic examinations: nodular appearing liver at abdominal imaging [ultrasound/computed tomography (CT)], reduced portal vein flow at ultrasound, increased liver stiffness, gastroesophageal varices at upper endoscopy; and 4) fibrosis stage 4 at histology. Exclusion criteria were: age >70 y, etiology of cirrhosis other than alcoholic, metabolic dysfunction-associated steatotic liver disease [22] or viral, alcohol consumption and/or active or recent (<6 mo) hepatitis C virus (HCV) and hepatitis B Virus (HBV) positive viremia, acute hepatic decompensation, hepatocellular carcinoma exceeding Milan for LT, diagnosis of extrahepatic neoplasms, previous LT, diagnosis of cardiac and/or respiratory failure. Fifteen healthy subjects and 25 patients with chronic hepatitis (CH) C were recruited as 2 different control groups. All recruited subjects were Caucasian and were invited to complete a 3-d food diary over the next 10 d. Of the 111 recruited, 96 patients with CIR completed the 3-d food diary. Of these 96 patients, n = 12 was excluded because they were lost to follow-up, and n = 4 due to LT within 6 mo of recruitment (Supplemental Figure 1). The study was approved by the ethics committee of Policlinico Umberto I (reference number 3420) and conducted in accordance with the Declaration of Helsinki and of Istanbul. Written informed consent was obtained from all participants.

Data collection

Demographic, nutritional, and clinical data were collected for each subject at the time of recruitment. All laboratory parameters were obtained using standard laboratory methods. In patients with CIR, the stage of disease was classified based on the MELD score, which was collected both at recruitment and 6 mo later to investigate possible temporal changes in liver disease severity. Changes in the MELD score over time were then expressed as DELTA-MELD, calculated as: [(MELD 6 mo after enrollment—MELD at enrollment)/time in days] × 180. DELTA-MELD values >0 were considered as stable/worsened MELD, whereas values <0 as improved MELD. Episodes of overt HE according to the West-Haven criteria were also recorded [23].

Nutritional characteristics

For each enrolled patient, a subjective global nutritional assessment (SGA) was carried out, which allowed each subject of the study population to be classified as well-nourished (SGA A), moderately malnourished (SGA B), or severely malnourished (SGA C) [24]. To avoid interoperator differences, the evaluations were carried out by the same operator. Traditional anthropometric measurements were also carried out. BMI was computed as weight expressed in kilogram divided by the height squared expressed in meters (kg/m²). Mid-arm circumference and triceps skinfold thickness were assessed by the same experienced clinician. The Harpenden skinfold caliper (John Bull British Indicators Ltd) was used to obtain triceps skinfold thickness measurements, which were performed following recommended protocols [25].

Daily fiber consumption

A 3-d food diary completed on 2 nonconsecutive weekdays and 1 weekend day within 10 d of enrollment was used to estimate daily nutrient and fiber dietary intakes [26,27]. To minimize errors, a trained dietitian provided each subject with detailed instructions at recruitment and then reviewed the questionnaires after completion during a face-to-face interview using a food atlas as a visual support to obtain an accurate and detailed report of dietary habits [28]. At enrollment, patients with CIR did not receive any dietary interventions other than current nutritional recommendations for cirrhosis [29], and all subjects were asked not to change their dietary habits during the study period. The 3-d food diaries were analyzed through the food composition database for epidemiologic studies in Italy (Food Composition Data Bank for Epidemiological Studies in Italy—BDA) (European Institute of Oncology, 2015) [30,31]. The nutritional data obtained were counted as the median daily intake for the 3 d covered by the diary. Since, to date, the food composition database for epidemiologic studies in Italy does not provide estimates of the food content of soluble and insoluble dietary fiber, the FiberTAG Repertoire of food products containing dietary fiber by Neyrinck et al. [32] has been used to supplement. Through this database, it was, in fact, possible to extrapolate not only the total fiber contained in foods but also the percentage of soluble and insoluble fiber. Daily fiber and nutrient intake was expressed in grams (g) or milligrams (mg) per body weight.

Porto-systemic shunt assessment

All abdominal CT scans realized for the local standard LT evaluation protocol within 30 d of enrollment were reviewed by radiologists with expertise in liver disease and portal hypertension. The CT scans were screened to assess the presence of any PS, spontaneous or iatrogenic. Spontaneous PSs were considered present in case of any nonphysiologic extrahepatic vein within the portal venous system or between the splanchnic veins and the systemic venous system. Each spontaneous PS analysis was performed by scrolling through the postcontrast multidetector CT acquisition 70 s after contrast agent injection (portal venous phase) in the axial plane, with a 3 mm slice thickness or, if available, 1 mm slice thickness. The radiologist looked for any additional veins leaving the inferior vena cava and iliac veins, portal vein, splenic vein, right/left renal vein, and superior/inferior mesenteric vein. The minimum diameter of spontaneous PSs that was considered was 5 mm, with complete contrast opacification of lumen vessels in the portal venous phase [33]. Gastroesophageal varices, when present, were not taken into consideration. In addition to spontaneous PSs, iatrogenic shunts were also recorded, such as the presence of transjugular intrahepatic PSs (TIPS).

Aims of the study

The primary outcome of the study was to investigate the effect of dietary fiber content on changes in liver disease severity in a cohort of patients with CIR. Changes in liver disease severity were considered as an improvement compared with stability/worsening of MELD score 6 mo after enrollment (DELTA-MELD). Secondary outcomes were to assess whether the effect of dietary fiber differed according to the presence of PSs and whether there was an association between serum ALT and ALP values.

Statistical analysis

The Kolmogorov-Smirnov test was used to assess the distribution of variables, showing that all continuous variables included in the study did not follow a normal distribution. Therefore, continuous variables were presented as medians and IQRs. Categorical variables were presented as numbers and percentages. To examine group differences, the Mann-Whitney U test and the χ² test were used as appropriate. To investigate possible independent associations between dietary fiber intake and specific clinical outcomes in patients with CIR, we constructed multivariate binary logistic regression models, considering as dependent variable 1) improvement in cirrhosis prognosis over time, indicated by DELTA-MELD scores <0, compared with stable or worsening disease and 2) presence or absence of cholestatic liver injury, defined by ALP values above or below the normal range 3) high or low consumption of soluble fiber, using 7 g/d as a cutoff. Covariate selection aimed to improve the completeness of our analysis while reducing the risk of type I error in our binary logistic regression. We included clinically significant variables (age and sex) as covariates regardless of their statistical significance in univariate analyses and variables with a P value <0.2 in univariate analyses. This approach was chosen to ensure that potentially relevant predictors were not excluded prematurely. Model selection was refined using backward stepwise elimination to ensure that only covariates with significant associations were included in the final model, combined with bootstrapping to improve robustness and predictive validity. The presence of gross residual confounding was excluded because the results of the logistic regression analyses did not change by including age as a covariate divided into 5 strata instead of as a continuous variable [34]. A test for collinearity was performed between all covariates considered in the different models, confirming that no significant collinearity was present, except, as expected, for different types of fiber (total, soluble, and insoluble), which were then entered separately in the different models. All collinearity tolerance coefficients were greater than the cutoff of 0.1, below which collinearity is present. We verified that all continuous independent variables satisfied linearity with their respective log odds using the Box-Tidwell transformation test. We tested, for each independent variable, for the presence of extreme outliers by calculating the cook distance. We found no substantial differences compared to the logistic regression results when repeated after removing the very few outliers found. There were no missing data, and no imputation methods were used. As a pilot investigation, a formal sample size calculation was not feasible due to the lack of data on this specific topic. The Hosmer-Lemeshow goodness-of-fit test was used to exclude a poor fit in all multivariate models. All statistical tests were 2-tailed, with a significance level set at P < 0.05. Data analyses were conducted using SPSS version 25.0 for Windows software (SPSS Inc), ensuring rigorous handling and accurate interpretation of statistical results.

Results

Dietary fiber intake in patients with CIR compared with that in patients with CH and healthy controls

Table 1 shows the demographic, clinical, nutritional, and dietary intake characteristics of patients with liver cirrhosis, those with CH, and healthy controls. There were no intergroup differences in age. Patients with CIR were more frequently male than subjects with CH. BMI and the frequency of overweight/obesity were higher in patients with CIR than in the other 2 groups. The frequency of diabetes, dyslipidemia, and arterial hypertension was 26%, 16%, and 24%, respectively, in patients with CIR but equal to 0 in the other 2 groups. Regarding nutritional assessment, patients with CIR appeared globally well nourished, with a median dry body weight of 80.1 (IQR: 68.0–93.2) kg. Only 6 patients with CIR were classified as SGA B and 1 as SGA C. No intergroup differences were present for the mid-arm circumference values and triceps skinfold measurements. Regarding daily food consumption, we found that, compared to healthy controls, patients with CIR ate significantly fewer total calories, proteins, and fats but equal amounts of carbohydrates, total, soluble, and insoluble fiber. Furthermore, there were no significant differences regarding the aforementioned food consumption between patients with CIR and those with CH. The percentage of patients with CIR consuming the minimum recommended amount for the general population of ≥25 g/d of total fiber was 3.8%, and only 1.3% consumed >30 g/d.

TABLE 1.

Demographic, clinical, nutritional, and dietary intake characteristics of patients with cirrhosis and chronic hepatitis and healthy controls.

	Patients with cirrhosis (n = 80)	Patients with chronic hepatitis (n = 25)	Healthy controls (n = 15)	P value cirrhotic vs. patients with chronic hepatitis	P value patients with cirrhosis vs. healthy controls
	Median (IQR) or number (%)
Age, y	59.5 (53.3–63.0)	59.0 (52.6–70.8)	58.0 (51.0–63.0)	0.158	0.810
Sex, M	70 (87.5)	17 (68.0)	12 (80)	0.024	0.426
BMI, (kg/m2),	27.5 (24.6–31.3)	24.6 (23.3–27.6)	24.9 (23.2–27.0)	0.006	0.017
Overweight/obesity	59 (73.8)	11 (44.0)	7 (46.7)	0.006	0.037
Diabetes	21 (26.3)	0 (0)	0 (0)	<0.001	<0.001
Dyslipidaemia	13 (16.3)	0 (0)	0 (0)	<0.001	<0.001
Arterial hypertension	19 (23.8)	0 (0)	0 (0)	<0.001	<0.001
Nutritional assessment:				0.491	0.650
SGA A	73 (91.3)	25 (100)	15 (100)
SGA B	6 (7.5)	0 (0)	0 (0)
SGA C	1 (1.3)	0 (0)	0 (0)
MAC, cm	30.0 (27.0–35.0)	29.3 (28.0–31.9)	32.0 (30.0–34.0)	0.582	0.417
Triceps skinfold measurements, mm	13.6 (7.7–20.1)	13.6 (11.8–15.9)	9.9 (7.5–14.9)	0.972	0.111
Energy, kcal/kg/d	20.8 (17.3–27.1)	23.5 (17.7–28.9)	25.8 (24.5–33.5)	0.232	0.005
Proteins dietary daily intake, g/kg of body weight	0.86 (0.66–1.11)	0.94 (0.78–1.12)	1.08 (0.86–1.41)	0.254	0.017
Fats dietary daily intake, g/kg of body weight	0.73 (0.56–1.06)	0.95 (0.66–1.14)	1.06 (0.81–1.29)	0.062	0.002
Carbohydrates dietary daily intake, g/kg of body weight	2.91 (2.24–3.68)	2.96 (2.33–3.53)	3.10 (2.64–3.80)	0.897	0.401
Total fiber dietary daily intake, mg/kg of body weight	198.6 (152.9–253.2)	240.9 (143.7–314.4)	190.9 (164.5–240.5)	0.089	0.841
Soluble fiber dietary daily intake, mg/kg of body weight	67.8 (52.6–88.9)	89.1 (50.8–107)	69 (56.1–78.3)	0.978	0.978
Insoluble fiber dietary daily intake, mg/kg of body weight	125 (97.4–168.4)	146.9 (93–196.6)	126 (104.8–158.2)	0.169	0.778

All data refer to enrollment unless differently specified.

Statistical tests: Mann-Whitney test for continuous variables and χ² for categorical variables.

All statistical tests were 2-tailed, with a significance level set at P < 0.05.

Abbreviations: BMI, body mass index; IQR, interquartile range; MAC, mid-arm circumference; SGA, subjective global nutritional assessment.

A low daily intake of fiber in the diet is associated with the improvement of the MELD score of patients with CIR over time

Among the 80 patients with CIR, the MELD score measured after 6 mo, compared with that measured at enrollment, was higher, equal, and lower in 45 (56%), 6 (8%), and 29 (36%) patients, respectively. Of the patients who had worsened at the second MELD score, 23 had increased MELD by ≥2 points. Of the patients who had improved, 8 had decreased MELD by ≥2 points.

We therefore wanted to verify whether, in the entire population of patients with CIR in the study, the change in the MELD score during a 6-mo period was associated with dietary fiber intake. To do this, we divided the patients with CIR into 2 subgroups: 1 with stable or worsening disease (i.e., the MELD score was unchanged or increased and therefore DELTA-MELD ≥0; n = 51) during the 6 mo of observation, and the other that had improved (i.e., the MELD score was decreased and thus DELTA-MELD <0; n = 29).

Despite equal MELD values at enrollment in the 2 groups, in the group with stable or worsening disease after 6 mo, the MELD score at the end of the study was significantly higher than that of the group that had experienced clinical improvement (Supplemental Table 1).

Regarding dietary food consumption, however, we found that patients with MELD scores that improved over time showed lower median daily dietary intake per unit of body weight of total [179.35 (IQR: 121.60–214.02) compared with 205.75 (IQR: 167.38–265.66) mg/kg, P = 0.005], soluble [61.54 (IQR: 41.28–73.08) compared with 74.68 (IQR: 56.03–92.71) mg/kg, P = 0.001] and insoluble [119.96 (IQR: 82.02–143.82) compared with 132.82 (IQR: 106.19–176.32) mg/kg, P = 0.038] fiber (Figure 1). Protein and fat intake were similar in the 2 groups, whereas there was a nonsignificant trend for lower total carbohydrate intake in the clinical improvement group (Supplemental Table 1).

FIGURE 1.

Boxplots of the amounts of fiber per kg of body weight consumed each day according to temporal changes in the MELD score by patients with cirrhosis. (A) Total fiber; (B) soluble fiber; (C) insoluble fiber. MELD, model for end-stage liver disease.

We, therefore, wanted to verify if, within the entire population of patients with CIR under study, there were any independent associations between the change in the MELD score over time and the fiber dietary consumption and whether these associations varied according to the amount of fiber consumed. We built separate multivariable binary logistic regression models for each fiber type, entering as a dependent variable in all models the improvement of the MELD score over time compared with its stability or worsening. After adjusting each of the models for age, sex, and arterial hypertension, daily consumption of total fiber [odds ratio (OR): 0.986; 95% CI: 0.977, 0.995, P = 0.001], of soluble fibers [OR: 0.960; 95% CI: 0.936, 0.984, P = 0.001] and insoluble fibers [OR: 0.983; 95% CI: 0.970, 0.996; P = 0.012] were independently and inversely associated with improvement in MELD score after 6 mo. All models showed significant values in the Hosmer-Lemeshow goodness-of-fit test, which allows us to exclude their poor fit (Supplemental Table 2). Finally, given the strong association between soluble fiber consumption expressed as a continuous variable in milligrams per kilogram (mg/kg) body weight and change in MELD score over time, we analyzed whether the association remained for soluble fiber consumption >7 g/d. In our cohort, 16 patients consumed >7 g of soluble fiber per day (Supplemental Table 3). These patients with high soluble fiber consumption did not differ from the other patients with CIR for demographic and clinical variables, with the exception of changes in the MELD score over time. In particular, the DELTA-MELD was significantly higher (P = 0.006), and the frequency of patients with stable or worsened MELD was significantly higher (P = 0.007) in the group with high soluble fiber consumption, being 36/16 (94%) in the latter and 36/64 (56%) in the low soluble fiber group. As shown in Figure 2, it is noteworthy that soluble fiber consumption >7 g/d was independently associated (OR: 11.667; 95% CI: 1.452, 93.724; P = 0.019) with stable or worsening MELD score in the overall cohort after adjustment for age and sex. The significance values of the Hosmer-Lemeshow statistic in this model were 0.651 and 0.509 in the first and second elimination steps of the stepwise backward analysis, respectively. These values ruled out poor model fit.

FIGURE 2.

Multivariate Binary Logistic Regression predicting the probability of consumption of soluble fiber higher than 7g/d adjusted for age and sex. MELD, model for end-stage liver disease; SPSS.

Different beneficial effects of a diet low in soluble or insoluble fiber on temporal changes in the MELD score of patients with CIR depending on whether PSs are present or absent

We then assessed whether the association between dietary fiber intake and temporal changes in the MELD score of patients with CIR differed according to the presence or absence of PSs. Forty-three (53.8%) patients with CIR had spontaneous PSs, of which 2 also had TIPS. Five patients had TIPS without spontaneous PSs. Together, we considered the 48 patients who had ≥1 PS and/or TIPS and compared them to 32 patients without PSs. Patients with CIR with and without PSs did not differ with respect to age, BMI, sex, nutritional status, comorbidities, and fiber dietary intake (Supplemental Table 4). Patients with CIR with PSs, compared with those without, had statistically significantly higher MELD scores at enrollment [13.1 (IQR: 11.5–15.4) compared with 11.3 (IQR: 8.3–14.2), P = 0.029] and, interestingly, higher ALP values [167.5 (IQR: 110.8–262.0) compared with 131.0 (IQR: 90.5–151.8) IU/L, P = 0.029], whereas serum aminotransferases did not differ in the 2 groups. During follow-up, no difference in DELTA-MELD was found between the group with PSs and that without [0.258 (IQR: –0.628 to 2.946) compared with 0.360 (IQR: –0.339 to 1.964), P = 0.767] The distribution of patients who improved the MELD score did not differ between the 2 groups, being 19/48 (39.6%) in the group with PSs and 10/32 (31.3%) in the no-shunt group (P = 0.448).

We then separately analyzed the group of patients with CIR with and without PSs. Within each of the 2 groups, no differences were found at enrollment for demographic, clinical, nutritional variables, and dietary intake of protein, fat, carbohydrate, and energy when comparing patients who had improved MELD scores over time with those without improvement (Table 2). However, within the group of patients with CIR with PSs, those who improved their MELD score during follow-up consumed significantly less soluble fiber (mg/kg/d) than patients who did not improve [61.8 (IQR: 45–71.1) compared with 74.7 (IQR: 59.3–108.7), P = 0.01), whereas no intergroup differences were found for consumption of total and insoluble fiber (Figure 3). Unlike patients with CIR with PSs, those without shunts who improved their MELD score during follow-up consumed significantly less total [146.8 (IQR: 92.8–215.3) compared with 191.3 (IQR: 167.3–272.7), P = 0.03] and insoluble fiber [94.5 (IQR: 72.9–132.7) compared with 121.2 (IQR: 105.9–181.8), P = 0.02)] expressed as milligrams per kilogram (mg/kg) for day than patients who did not improve, whereas soluble fiber consumption did not differ (Figure 4).

TABLE 2.

Demographic, clinical, nutritional, and dietary intake characteristics of patients with cirrhosis subgrouped according to the presence or absence of portosystemic shunts and to their model for end-stage liver disease score changes over time.

	Patient with cirrhosis without portosystemic shunts			Patients with cirrhosis with portosystemic shunts
	Stable or worsened MELD score (n = 22)	Improved MELD score (n = 10)	P value	Stable or worsened MELD score (n = 29)	Improved MELD score (n = 19)	P value
	Median (IQR) or n (%)			Median (IQR) or n (%)
Age, y	61.5 (58–64.3)	60.5 (52.5–63)	0.535	58 (54–62.5)	56 (51.5–62.5)	0.263
Sex, M	20 (90.9)	7 (70)	0.293	27 (93.1)	16 (84.2)	0.372
BMI	26.8 (23.9–30.8)	26.2 (22.9–30.6)	0.589	27.9 (24.5–31.9)	27.4 (25.5–30.7)	0.891
Overweight/obesity	14 (63.6)	7 (70)	1.000	22 (75.9)	16 (84.2)	0.719
Diabetes	6 (27.3)	1 (10)	0.387	8 (27.6)	6 (31.6)	0.766
Dyslipidaemia	4 (18.2)	1 (10)	1.000	5 (17.2)	3 (15.8)	1.000
Arterial hypertension	7 (31.8)	1 (10)	0.380	9 (31)	2 (10.5)	0.161
Cirrhosis etiology:
Viral	12 (54.5)	8 (80)	0.248	12 (41.4)	6 (31.6)	0.493
Alcohol	11 (50)	6 (60)	0.712	14 (48.3)	12 (63.2)	0.312
MASLD	18 (81.8)	8 (80)	1.000	23 (79.3)	17 (89.5)	0.451
Ascites	12 (54.5)	5 (50)	0.811	12 (41.4)	6 (31.6)	0.493
Previous hepatic encephalopathy	8 (36.4)	3 (30)	1.000	17 (58.6)	10 (52.6)	0.683
HCC	13 (59)	6 (60)	1.000	11 (37.9)	5 (26.3)	0.535
Serum AST, UI/L	36 (27.5–78.3)	37.5 (26.5–91.3)	0.984	50 (35.5–73.5)	53 (37–69)	0.941
Serum ALT, UI/L	32 (18–66)	30.5 (18.5–67.8)	0.952	34 (22–64.5)	33 (25–43.5)	0.658
Serum ALP, UI/L	131 (91.5–154)	130.5 (79.3–148.8)	0.747	157 (120–250)	168 (85.5–310.5)	0.650
MELD at baseline	11.2 (7.9–14.8)	11.4 (9.1–12.4)	0.764	12.9 (11.4–15)	13.8 (11.6–16.6)	0.356
MELD after 6 mo	13.5 (9.5–17.3)	10.5 (7.8–11.3)	0.003	15 (14–19)	12.9 (14.5)	0.002
Δ-MELD	1.07 (0.3–2.6)	–0.8 (–1.52 to –0.34)	<0.001	1.8 (0.62–3.7)	–0.8 (–2.5 to –0.4)	< 0.001
Charlson modified comorbidity index:			0.167			0.765
0	10 (45.5)	8 (80)		10 (34.5)	9 (47.4)
1–2	9 (40.9)	1 (10)		16 (55.2)	8 (42.1)
>2	3 (13.6)	1 (10)		3 (10.3)	2 (10.5)
Nutritional assessment:			1.000			0.286
SGA A	20 (90.9)	10 (100)		27 (93.1)	16 (84.2)
SGA B	2 (9)	0 (0)		1 (3.4)	3 (15.8)
SGA C	0 (0)	0 (0)		1 (3.4)	0 (0)
MAC, (cm)	29 (26.8–33)	27.8 (23.8–35.5)	0.535	31 (27–34.5)	31 (27–36)	0.531
Triceps skinfold measurements (mm)	13.5 (8.5–18.7)	11.7 (5.7–21.9)	0.617	14.8 (9–20.3)	11.8 (7.1–18.8)	0.439
Energy, kcal/kg/d	26.7 (18.6–30.9)	19.4 (14.9–23.9)	0.108	20.5 (17.1–26.5)	20.7 (14.5–24.3)	0.650
Proteins dietary daily intake (g/kg of body weight)	1 (0.7–1.2)	0.7 (0.6–1)	0.128	0.8 (0.7–1.1)	0.8 (0.6–1)	0.910
Fats dietary daily intake (g/kg of body weight)	0.9 (0.7–1.2)	0.7 (0.6–1.05)	0.246	0.6 (0.56–0.9)	0.7 (0.5–0.9)	0.858
Carbohydrates dietary daily intake (g/kg of body weight)	3.6 (2.4–4.2)	2.3 (1.9–3.2)	0.094	3.2 (2.4–3.6)	2.7 (2.2–3.2)	0.393

All data refer to enrollment unless differently specified.

Statistical tests: Mann-Whitney test for continuous variables and χ² for categorical variables.

All statistical tests were 2-tailed, with a significance level set at P < 0.05.

Abbreviations: ALP, alkaline phosphatase; ALT, alanine transaminase; AST, aspartate transaminase; BMI, body mass index; HCC, hepatocellular carcinoma; IQR, interquartile range; MAC, mid-arm circumference; MASLD, Metabolic dysfunction-associated steatotic liver disease; MELD, model for end-stage liver disease; SGA, subjective global nutritional assessment.

FIGURE 3.

Boxplots of the amounts of fiber per kilogram of body weight consumed each day according to temporal changes in the MELD score by patients with cirrhosis with portosystemic shunts. (A) Total fiber; (B) soluble fiber; (C) insoluble fiber. MELD, model for end-stage liver disease.

FIGURE 4.

Boxplots of the amounts of fiber per kilogram of body weight consumed each day according to temporal changes in the MELD score by patients with cirrhosis without portosystemic shunts. (A) Total fiber; (B) soluble fiber; (C) insoluble fiber.

Table 3 shows the unadjusted and adjusted binary logistic regression analyses regarding the associations between dietary fiber consumption and the change in MELD score at the end of the follow-up, performed separately in the group of patients with CIR with or without PSs. After adjustment for confounding factors, improvement in MELD score after 6 mo was inversely and independently associated with daily total fiber consumption in both groups of patients with CIR with and without PSs. Regarding fiber subtypes, however, in patients with CIR with PSs, MELD improvement was strongly associated with low soluble fiber consumption, whereas insoluble fiber consumption was not associated with clinical outcomes. In contrast, in patients with CIR without shunts, improvement in MELD was associated with low consumption of insoluble fiber, whereas there was a nonsignificant trend regarding low consumption of soluble fiber. The significance values of the Hosmer-Lemeshow statistic in the various models excluded their poor fit (Supplemental Table 2). In our cohort, soluble fiber intake >7 g/d was present in 11 and 5 patients with CIR in the PSs and non-PSs groups, respectively. As shown in Figure 2, soluble fiber intake >7 g/d was independently associated (OR: 9.474; 95% CI: 1.099, 81.684; P = 0.023) with stable or worsening MELD score in patients with PSs. The significance values of the Hosmer-Lemeshow statistic in this model were 0.763 and 0.843 in the first and second elimination steps of the stepwise backward analysis, respectively. These values ruled out poor model fit. The association between soluble fiber intake >7 g/d and the time course of MELD was not statistically significant in patients without PSs (data not shown).

TABLE 3.

Unadjusted and adjusted results of logistic regression analysis predicting the probability of model for end-stage liver disease score amelioration over time and fiber consumption in patients with cirrhosis according to the presence or absence of portosystemic shunts.

	Daily dietary intake	Unadjusted analysis				Adjusted analysis
		OR	95% CI		P value	OR	95% CI		P value
			Lower	Upper			Lower	Upper
Patients with cirrhosis and without portosystemic shunts	Total fiber intake (mg/kg of body weight)	0.983	0.968	0.999	0.032	0.981	0.965	0.998	0.005
	Soluble fiber intake (mg/kg of body weight)	0.967	0.932	1.003	0.074	0.966	0.930	1.004	0.050
	Insoluble fiber intake (mg/kg of body weight)	0.974	0.951	0.999	0.038	0.968	0.939	0.997	0.005
Patients with cirrhosis with portosystemic shunts	Total fiber intake (mg/kg of body weight)	0.991	0.9,82	1.001	0.067	0.985	0.974	0.997	0.016
	Soluble fiber intake (mg/kg of body weight)	0.962	0.934	0.991	0.010	0.946	0.912	0.982	0.001
	Insoluble fiber intake (mg/kg of body weight)	0.993	0.981	1.005	0.272	0.990	0.977	1.003	0.152

All multiple regression models were adjusted for age, sex, and arterial hypertension.

Statistical tests: Univariate (unadjusted analysis) and multivariate (adjusted) binary logistic regression.

All statistical tests were 2-tailed, with a significance level set at P < 0.05.

Abbreviations: CI, confidence interval; OR, odds ratio.

A diet low in soluble fiber is associated with less cholestatic damage in patients with CIR with PSs but with greater hepatocyte damage in patients with CH

To test whether there was an association between fiber consumption and serum markers of cholestatic damage, we compared fiber consumption in patients with serum ALP values within the normal range compared with those with elevated values at enrollment. Among patients with CIR with PSs, those with normal ALP (n = 19), compared with those with abnormal values (n = 29), consumed significantly less soluble fiber [62.7 mg/kg (IQR: 54.4–72.4) compared with 74.7 mg/kg (IQR: 60.8–108.7), P = 0.03] (Figure 5). Interestingly, among the 11 patients with CIR with PSs who consumed >7 g/d of soluble fiber, 10 had abnormal serum ALP values at enrollment and subsequently had worsened MELD scores after 6 mo. The 1 patient who consumed >7 g/d of soluble fiber and who had a normal ALP at enrollment also had a mild improvement in MELD at follow-up. This patient weighed 131 kg and, therefore, had a low weight-adjusted intake of soluble fiber.

FIGURE 5.

Boxplots of the amounts of fiber per kilogram of body weight consumed each day according to serum ALP activity at enrollment by patients with cirrhosis with portosystemic shunts. (A) Total fiber; B) soluble fiber; C) insoluble fiber. ALP, alkaline phosphatase.

We therefore wanted to verify whether, within the population of patients with CIR with PSs, the association between low consumption of soluble fiber and normal serum ALP activity was independent of confounding factors. Within the group of patients with CIR with PSs, no demographic, clinical, and nutritional differences were found between the group that had normal ALP values and those with abnormal values (Supplemental Table 5). In binary logistic analysis, using normal or abnormal serum ALP as the dependent variable and age, sex, BMI, and viral etiology of cirrhosis as covariates, daily soluble fiber consumption was independently and inversely associated with normal ALP (OR: 0.964; 95% CI: 0.936, 0.993; P = 0.010). The significance values of the Hosmer-Lemeshow statistic in this model were 0.536, 0.642, 0.681, and 0.353 in the first, second, third, and fourth elimination steps of the stepwise backward analysis, respectively. These values allowed us to exclude a poor fit of the model.

Analyzing patients with CIR without PSs, we found no differences in total, soluble, or insoluble fiber consumption between patients with normal (n = 16) or abnormal ALP (n = 16) (Supplemental Figure 2). In patients with CH, only 3 had abnormal ALP values, and no differences were found in fiber consumption compared with patients with normal ALP values (Supplemental Figure 3).

To test whether there was an association between fiber consumption and serum markers of hepatocellular injury, we compared fiber consumption in patients with normal or abnormal serum ALT values at enrollment. No difference was found in the consumption of total, soluble, or insoluble fiber in the group of patients with CIR with PSs, comparing those with abnormal ALT (n = 13) and those with normal ALT (n = 35) (Supplemental Figure 4). No difference in fiber consumption between patients with abnormal ALT (n = 10) and those with normal ALT (n = 22) was also present in the group of CIRs without PSs (Supplemental Figure 5).

Regarding patients with CH, those with abnormal ALT (n = 15) compared with those with normal ALT (n = 10) showed a significantly lower daily consumption of soluble fiber, whereas no difference was found for total and insoluble fiber consumption (Figure 6). Furthermore, patients with CH and abnormal ALT compared with those with normal ALT had significantly higher triceps skinfold values, whereas there were no differences in BMI, age, sex, and blood HCV RNA (Supplemental Table 6).

FIGURE 6.

Boxplots of the amounts of fiber per kilogram of body weight consumed each day according to serum ALT activity at enrollment by chronic hepatitis. (A) Total fiber; (B) soluble fiber; (C) insoluble fiber. ALT, alanine transaminase.

Discussion

This study focused on the role of dietary fiber consumption in decompensated CIRs outpatients being evaluated for LT. For this purpose, at enrollment, we measured their total intake of soluble and insoluble fiber, as well as their MELD score and blood indices of hepatocellular and cholangiocyte damage [20,21]. Six months later, during which patients had been advised not to change their eating habits, we re-evaluated the MELD score to compare it to that at enrollment.

Analyzing the entire cohort of patients with CIR, we found that the improvement in the MELD score after 6 mo was associated, after adjustment for confounding factors, with lower consumption of insoluble and, more strongly, soluble fiber. We then analyzed patients with CIR and with PS separately from those without. As expected and in line with the known association of PS with more decompensated cirrhosis [[16], [19]], patients with PS had a higher baseline MELD score than those without PS. However, changes in cirrhosis severity 6 mo after enrollment did not differ between patients with or without PS. We found that improvement in MELD over time was independently associated with lower insoluble fiber intake in patients without PS and soluble fiber intake in patients with PS. Interestingly, the odds of having a stable or worsening MELD score over time in patients with a daily soluble fiber intake >7 g/d, compared with those with a lower intake, was 11 times higher in the entire CIR population and 9 times higher in those with PS.

In patients with PSs, an independent association was also found between normal serum ALP values at the time of enrollment and low consumption of soluble, but not insoluble, fiber. Finally, we found that in patients with CH, and therefore less advanced disease than cirrhosis, those with normal serum ALT concentrations at enrollment consumed more soluble fiber than those with abnormal ALT. Therefore, regarding soluble fiber consumption, we found contrasting associations with clinical outcomes, depending on the stage of chronic liver disease: beneficial in CH but harmful in cirrhosis with PSs, which we found to characterize more advanced cirrhosis than the 1 without shunts.

Although high-fiber diets are generally considered beneficial, recent data demonstrate that the effect on the liver of metabolites produced by the gut microbiota during dietary supplementation with soluble fiber, such as short-chain fatty acids, may depend on the context [[9], [14]]. In fact, short-term intake of soluble fibers can cause hepatocyte damage in some healthy subjects, and prolonged intake causes hepatocyte damage, cholestasis, and fibrosis in animal models in the presence of intestinal dysbiosis and high serum bile acid concentration [11,13,14]. Although in the present study, we did not investigate the gut microbiota, bile acids, and short-chain fatty acids, we hypothesize that the different effects of soluble fiber consumption that we found depending on the stage of chronic liver disease could be caused by the well-demonstrated peculiarities of the intestinal microbiota in different stages [[35], [36], [37], [38], [39], [40]].

The characteristic microbiota of cirrhosis with PSs, which is known to be more advanced than that without shunts, could produce high quantities of soluble fiber metabolites that are hepatotoxic [16,18,19]. Furthermore, the presence of PSs could represent a further element associated with a negative clinical response to soluble fibers regardless of the severity of cirrhosis. PSs could, in fact, favor liver damage both by changing the intestinal microbiota and by increasing the spillover into the general circulation of hepatotoxic metabolites produced in the intestine by the interaction of soluble fibers with the altered microbiota. As regards the first hypothesis, although there are no data demonstrating peculiar alterations of the intestinal microbiota induced by spontaneous PSs in patients with CIR, it is possible that this occurs in a similar way to what has been demonstrated prospectively after TIPS placement [41]. The second hypothesis is supported by the fact that some studies have reported that high serum concentrations of bile acids and negative hepatic clinical outcomes are associated with the presence of PSs not only acquired spontaneously in patients with CIR but also congenital ones [[14], [19],42]. Indeed, high serum concentrations of bile acids are always present both in humans and in animals with congenital PSs. As regards liver outcomes, in the C57BL/6 mouse, dietary supplementation with soluble fibers invariably causes severe chronic cholestatic liver damage with fibrosis only in the presence of congenital PSs [14]. However, no data are available on dietary fiber intake and intestinal microbiota that can explain why some, but not all, human subjects with congenital PSs present cholestasis and hepatic fibrosis [15,43].

In our study, in patients with CIR without PSs, we did not find an association between fiber consumption and cholestatic damage, whereas the improvement of the MELD score over time was associated with low consumption of insoluble fibers. As regards the lack of association between fiber consumption and cholestatic damage, it is conceivable that the latter occurs only in the presence of shunts, whether spontaneously acquired in CIRs or congenital, because they increase the contact of cholangiocytes with toxic metabolites of intestinal derivation as the blood supply of these cells is predominantly arterial from the peribiliary plexus [44]. The explanation for the detrimental effect of insoluble fiber consumption on MELD score changes is unclear.. One hypothesis is that, in the absence of PSs, the type of intestinal dysbiosis of cirrhosis is different from that in the presence of shunts and that the insoluble fibers select bacteria capable of producing harmful metabolites reaching the liver through the portal circulation.

The positive effect on the clinical outcomes of patients with CIR by keeping the consumption of dietary fiber particularly low that we found is in contrast with the results of a previous study performed on Iranian patients with predominantly autoimmune cirrhosis and another on Chinese patients with minimal HE [7,8]. This discrepancy could be explained by differences in terms of the composition of the intestinal microbiota and its ability to produce metabolites for geographic and dietary reasons or for the etiology of cirrhosis or the type of complications [[45], [46], [47]].

Nutritional guidelines and recommendations for patients with CIR do not specify the amount of fiber to consume daily [29]. In light of our data, fiber consumption should be individualized in patients with CIR, and patients with PSs should consume a lower amount of soluble fiber than what is recommended for the general population by some experts, i.e., 6–8 g/d [48]. Indeed, all but 1 of our patients with CIR with PSs consuming >7 g/d of soluble fiber had increased serum ALP at enrollment and worsened MELD scores at follow-up. Interestingly, the body weight of the 1 patient consuming >7 g/d of soluble fiber who had normal ALP values and a slight improvement in MELD over time was very high, making his fiber intake low once standardized for body weight.

In conclusion, our results show that, although patients with CIR consume similar amounts of dietary fiber to healthy subjects and patients with CH, consumption is very often lower than what is recommended for the general population. We found that, although in patients suffering from CH, relatively higher consumption of soluble fiber is associated with less hepatocyte damage, in patients with CIR with PSs, it is harmful. These findings have clinical relevance for patients with CIR in general and for LT candidates.

Although our data suggest an important role of dietary fiber content in liver cirrhosis, our study presents numerous limitations. First, from a methodological point of view, our study is monocentric, and the number of patients considered in the subanalyses of patients with CIR with or without PSs was relatively small. Furthermore, being an observational study, it may suffer from selection bias, information bias, and the failure to account for confounding variables. Although we verified the assumptions of our multivariate models and applied statistical corrections, the possibility of an increase in type I error due to multiple comparisons should be considered as a limitation of our study. Regarding the methodology of assessing eating habits with a 3-d food diary, although it has been clearly identified as the best tool to verify habits in patients with CIR, it has the known limitation of being self-reported, and there is the possibility of bias in its completion. To minimize bias, each diary was reviewed by an experienced dietitian. Furthermore, it was not possible to assess whether there had been any change in the patient’s diet over time during the study. We believe that future studies, possibly randomized, with larger samples, will be needed to validate our results. Another limitation of our study is that we categorized dietary fiber into soluble and insoluble. Although this classification is widely used, it presents significant limitations, because not all fibers can be distinctly categorized [32,49,50]. In fact, there is a general awareness of the need to find a more appropriate classification of fibers, which also takes into account the functional aspect [49,50]. Lastly, owing to the absence of a universally accepted dietary fiber database, we decided to employ a relatively recently published dataset in this study. This dataset was realized by reporting the median fiber content starting from the different databases [32]. In the future, it will be essential to develop and validate a food database for the Italian population that also takes into account a functional classification of fibers.

For these reasons, future studies should be designed to verify the effects of soluble and insoluble dietary fibers on the complex bidirectional relationship existing between the microbiota and bile acids or other bacterial metabolites capable of influencing the clinical outcomes of patients with CIR [51]. These studies will allow us to identify, from a personalized medicine and nutrition perspective, the subgroups of patients with CIR, also based on the presence or absence of PSs, in which the consumption of fibers and/or subtypes of fibers should be limited.

Author contributions

The authors’ responsibilities were as follows – SP: wrote the original draft of the manuscript, study design, and investigation; SC, ED, MC, PL, ADS: investigation and resources; FF, M Mischitelli: investigation; AC: formal analysis and methodology; DA, M Muscaritoli: resources and supervision; SGC: study design, formal analysis, data curation and writing the original and final version of the manuscript. All authors actively participated in the realization of the study and were involved with the writing of the manuscript; and all authors: read and approved the final manuscript.

Data availability

Data described in the manuscript, code book, and analytic code will be made available upon request to the corresponding author.

Funding

This study was supported by the “Ateneo Research Fund,” Sapienza University of Rome, and the supporting source had no restrictions regarding publication.

Conflict of interest

The authors report no conflicts of interest.

Appendix A. Supplementary data

The following is the Supplementary data to this article: