Cross-sectional association between gallstone disease and nonalcoholic fatty liver disease: Exploring the mediating roles of hematological and biochemical markers

Abstract

Nonalcoholic fatty liver disease (NAFLD) is the most common chronic liver disorder worldwide and is closely associated with metabolic comorbidities. Gallstone disease (GSD), another prevalent hepatobiliary condition, frequently coexists with NAFLD; however, their relationship and underlying mediating mechanisms remain unclear. This study aimed to investigate the association between a history of GSD and NAFLD using data from National Health and Nutrition Examination Survey 2017-March 2020. Restricted cubic spline models were employed to assess the linearity of this association, stratified analyses were conducted to evaluate the effect of cholecystectomy and sex, and least absolute shrinkage and selection operator regression combined with mediation analysis was used to screen for hematological and biochemical mediators. Among 3401 participants, 324 had GSD. Compared with those without GSD, affected individuals were older, more likely to be female, and had a higher prevalence of metabolic comorbidities. They also exhibited lower levels of albumin and iron, but higher levels of alkaline phosphatase (ALP), lactate dehydrogenase, white blood cell count (WBC), neutrophils, platelet count, red cell distribution width, and high-sensitivity C-reactive protein, along with reduced red blood cell count, hemoglobin, and hematocrit. The prevalence of NAFLD was significantly higher among participants with GSD (70.37% vs 53.85%). Multivariable logistic regression confirmed an independent association between GSD and NAFLD (odds ratio = 1.747, 95% confidence interval: 1.334–2.302). Restricted cubic spline analysis demonstrated a linear relationship, and the results were not materially modified by cholecystectomy or sex. Mediation analysis indicated that ALP, alanine aminotransferase (ALT), uric acid, and WBC significantly mediated the association, accounting for 6.08%, 19.54%, 11.37%, and 5.52% of the effect, respectively. GSD is independently associated with NAFLD, and this relationship is partly mediated by hematological and biochemical markers (ALP, alanine transaminase, uric acid, and WBC), suggesting shared metabolic and hepatobiliary pathways between the 2 conditions.

1. Introduction

Nonalcoholic fatty liver disease (NAFLD) has emerged as a global public health challenge, with an estimated prevalence of approximately 30% in the general adult population and up to 50% among individuals with metabolic risk factors such as obesity, type 2 diabetes mellitus, and dyslipidemia.^[1,2] NAFLD encompasses a spectrum of hepatic conditions ranging from nonalcoholic fatty liver to nonalcoholic steatohepatitis, which may progress to cirrhosis, hepatocellular carcinoma, and end-stage liver disease.^[1,3] Beyond liver-related morbidity and mortality, NAFLD is strongly associated with an increased risk of cardiovascular disease (CVD), the leading cause of death in affected individuals, as well as other metabolic complications, making it a major contributor to the global disease burden.^[4]

Gallstone disease (GSD), characterized by the formation of gallstones within the gallbladder, is a highly prevalent hepatobiliary disorder affecting approximately 10% to 20% of adults in the United States.^[5,6] Although most individuals with GSD remain asymptomatic, a substantial proportion develop symptomatic conditions such as biliary colic and cholecystitis, or even severe complications including cholangitis and pancreatitis.^[6,7] The primary treatment for GSD is cholecystectomy, which represents the most common abdominal surgical procedure worldwide.^[8] Similar to NAFLD, GSD shows strong epidemiological associations with metabolic risk factors, including obesity, type 2 diabetes mellitus, dyslipidemia, and insulin resistance.^[9,10] Clinical observations and epidemiological studies have frequently reported the co-occurrence of GSD and NAFLD, suggesting that these 2 diseases are not merely coincidental but may share common underlying biological mechanisms.^[11,12]

Despite the growing body of evidence supporting a GSD-NAFLD association, several critical questions remain unanswered. First, most previous studies have focused primarily on establishing the existence of this association rather than elucidating the potential mediating factors that might explain the link. Hematological and biochemical markers, such as liver enzymes, inflammatory indicators, and metabolic parameters, are often altered in both GSD and NAFLD and may act as intermediate steps in the causal pathway from GSD to NAFLD.^[13–19] Second, the role of cholecystectomy in the GSD-NAFLD association remains controversial. As cholecystectomy is the definitive treatment for GSD and given that GSD and cholecystectomy are almost causally linked, it is challenging to disentangle their independent effects on NAFLD. To date, no large-scale study has specifically assessed whether cholecystectomy modifies the association between GSD history and NAFLD through stratified analyses among GSD patients. Moreover, given the markedly higher prevalence of biliary diseases in women compared with men, it remains unclear whether the effects of GSD or cholecystectomy on NAFLD differ by sex. Third, the shape of the GSD-NAFLD association (linear or nonlinear) has not been characterized. While GSD is typically treated as a binary variable (present or absent), the relationship between GSD and NAFLD severity could follow a linear pattern. Restricted cubic spline (RCS) analysis has not yet been applied to this question, limiting our understanding of whether NAFLD risk increases proportionally with GSD.

The objectives of this study were to (1) investigate the cross-sectional association between GSD history and NAFLD in a large, representative sample while adjusting for multiple potential confounders; (2) evaluate the linearity of this association with the median controlled attenuation parameter (CAP, used for NAFLD diagnosis) through RCS analysis; (3) to clarify the impact of cholecystectomy on the GSD–NAFLD relationship by conducting stratified analyses among participants with a history of GSD, and further performing sex-stratified and interaction analyses to assess potential gender-specific modification; and (4) screen for specific hematological and biochemical markers mediating the GSD-NAFLD association using least absolute shrinkage and selection operator (LASSO) regression for variable selection, followed by formal mediation analysis.

2. Methods

2.1. Data source and study population

Data for this cross-sectional analysis were extracted from the National Health and Nutrition Examination Survey (NHANES), a continuous, cross-sectional survey conducted by the National Center for Health Statistics at the Centers for Disease Control and Prevention. Data were collected through household interviews (including demographics, medical history, and lifestyle factors) and physical examinations (including anthropometric measurements, laboratory testing, and imaging). For the present study, we included data from NHANES 2017-March 2020 Pre-pandemic. The study population was limited to adults aged 20 years or older. Participants were excluded if they: had missing data on GSD history or NAFLD status; had a history of alcoholic liver disease (defined as self-reported excessive alcohol consumption of ≥ 5 drinks per day for men or ≥ 4 drinks per day for women), viral hepatitis (hepatitis B or C), or were pregnant; had missing data on cholecystectomy status for stratified analyses; or had missing data on any hematological or biochemical markers for mediation analysis, or on any of the predefined covariates (Fig. 1).

Figure 1.

Flowchart for selecting participants from NHANES. GSD = gallstone disease, NAFLD = nonalcoholic fatty liver disease, NHANES = National Health and Nutrition Examination Survey.

2.2. Variable definitions

2.2.1. Outcome variable: NAFLD

NAFLD was diagnosed based on liver ultrasound transient elastography, defined as median CAP ≥248 dB/m, in the absence of: excessive alcohol consumption (≥5 drinks per day for men or ≥4 drinks per day for women, based on self-reported alcohol intake); viral hepatitis (hepatitis B surface antigen positivity or hepatitis C RNA positivity).^[20]

2.2.2. Exposure variable: GSD

GSD was defined as a self-reported, physician-diagnosed gallstone disease. Participants were asked: “Has a doctor or other health professional ever told you that you have gallstones?” Responses were coded as “yes” (GSD) or “no” (no GSD).

2.2.3. Mediator variables: hematological and biochemical markers

Potential mediators included 32 hematological and biochemical markers, all measured using standardized laboratory protocols in NHANES:

Hematological markers: Basophils number (×10⁹/L), eosinophils number (×10⁹/L), hematocrit (HCT, %), hemoglobin (HGB, g/dL), high-sensitivity C-reactive protein (hsCRP, mg/L), lymphocyte number (×10⁹/L), mean cell volume (fL), monocyte number (×10⁹/L), mean platelet volume (fL), neutrophil count (NEUT, ×10⁹/L), white blood cell count (WBC, ×10⁹/L), platelet count (PLT, ×10⁹/L), red blood cell count (RBC, ×10¹²/L), red cell distribution width (RDW, %).

Biochemical markers: Albumin (Alb, g/L), alkaline phosphatase (ALP, U/L), alanine transaminase (ALT, U/L), aspartate aminotransferase (U/L), bicarbonate (mmol/L), blood urea nitrogen (mg/dL), chloride (mmol/L), folate (nmol/L), gamma glutamyl transferase (U/L), globulin (g/L), iron (μmol/L), lactate dehydrogenase (LDH, U/L), phosphorus (mmol/L), potassium (mmol/L), serum creatinine (Scr, μmol/L), sodium (mmol/L), total bilirubin (μmol/L), uric acid (UA, μmol/L).

2.2.4. Covariates

Demographic variables: Age (years), gender (male, female), educational level (less than high school, high school graduate, higher than high school), marital status (married/living with partner, widowed/divorced/separated, never married), income-to-poverty ratio (PIR: <1.0 = low, 1.0–4.0 = middle, >4.0 = high, without data = unknown).

Anthropometric variable: A body shape index (ABSI), a measure of body shape that correlates more strongly with metabolic risk than body mass index.^[21] ABSI was calculated as: waist circumference (m)/(body mass index [kg/m²]^[2/3] × height [m]^[1/2]).

Lifestyle variable: Exercising status (yes, no; defined as engaged in at least 10 minutes of moderate or vigorous intensity activity per week), smoking status (yes, no; defined as had smoked at least 100 cigarettes in entire life).

Comorbidity variables: CVD (yes, no; defined as self-reported physician diagnosis of congestive heart failure, coronary heart disease, angina, heart attack, or stroke), diabetes mellitus (yes, no; defined as self-reported physician diagnosis of diabetes, taking insulin now, or taking diabetic pills to lower blood sugar, glycohemoglobin≥6.5%, glucose≥11.1 [mmol/L], or fasting glucose≥7.0 [mmol/L]), hypertension (yes, no; defined as self-reported physician diagnosis, use of antihypertensive medications, or a 3-time average systolic blood pressure ≥140 mm Hg or diastolic blood pressure ≥90 mm Hg), hyperlipidemia (yes, no; defined as self-reported physician diagnosis of high cholesterol level or use of lipid-lowering medications, high density lipoprotein≤1.03 [mmol/L] for men and≤1.29 [mmol/L] for women, triglyceride≥1.694 [mmol/L], total cholesterol≥5.17 [mmol/L], low density lipoprotein≥3.362 [mmol/L]), lung disease (yes, no; defined as self-reported physician diagnosis of emphysema, COPD or chronic bronchitis), kidney stone (yes, no; defined as self-reported physician diagnosis of kidney stones).

Cholecystectomy status (yes, no; defined as self-reported history of gallbladder removal) was not included in the main multivariable model because it is strongly causally related to GSD, being performed almost exclusively for this condition, and its inclusion could introduce collinearity and bias the estimate of the GSD-NAFLD association. Instead, cholecystectomy was evaluated in stratified analyses among GSD patients.

2.3. Statistical analysis

2.3.1. Descriptive statistics

Normality tests were conducted on all continuous variables, and only ABSI showed a normal distribution (Fig. S1, Supplemental Digital Content, https://links.lww.com/MD/R402), therefore, for statistical convenience, all continuous variables were expressed as median and interquartile range and compared between participants with and without GSD using the Wilcoxon rank-sum test, while categorical variables were presented as counts and percentages and compared using Pearson chi-square test. Collinearity of hematological and biochemical markers was detected using the Spearman correlation coefficient matrix (Fig. S2, Supplemental Digital Content, https://links.lww.com/MD/R402).

2.3.2. Multivariable logistic regression

Multivariable logistic regression models were constructed to evaluate the association between GSD and NAFLD. Three models were specified: Model 1 was unadjusted, with no covariates included; Model 2 was partially adjusted for age, gender, marital status, education level, and PIR; and Model 3 was fully adjusted for all predefined covariates, including age, gender, marital status, education level, PIR, exercising status, smoking status, ABSI, diabetes mellitus, hypertension, hyperlipidemia, CVD, lung disease, and kidney stone.

2.3.3. RCS analysis

RCS analysis was used to evaluate the shape of the association between GSD (treated as a binary variable: no GSD, GSD) and liver ultrasound-derived parameters (median CAP). The RCS model was fitted with 3 knots to balance flexibility and stability. The model was adjusted for all covariates in Model 3. The linearity of the association was tested by comparing the fit of the RCS model with a linear model using the Akaike information criterion and by assessing the significance of the nonlinear term in the RCS model. A nonsignificant nonlinear term (P >.05) indicated an approximately linear association.

2.3.4. Stratified analysis by cholecystectomy and sex

To evaluate the impact of cholecystectomy on the GSD–NAFLD association, participants with GSD were stratified by cholecystectomy status, and multivariable logistic regression models adjusted for all covariates in Model 3 were applied. Sex-stratified analyses and interaction tests were additionally performed in both the overall population and the GSD subgroup to examine potential gender-related effect modification.

2.3.5. LASSO regression for mediator selection and mediation analysis

Given the large number of potential mediators (32 hematological and biochemical markers), LASSO regression was used to select markers most strongly associated with NAFLD, reducing the risk of overfitting and multicollinearity in subsequent mediation analyses. The LASSO model was fitted with NAFLD as the binary outcome, all 32 markers as predictors, and all covariates in Model 3 as adjustment variables. The optimal lambda (λ) value was selected using 10-fold cross-validation. Two λ values were considered in the LASSO regression: λ_min, representing the λ value with the minimum cross-validation error, and λ_1se, the largest λ value within 1 standard error of λ_min, which serves as a more stringent criterion that selects fewer variables. Mediation analysis was subsequently conducted to evaluate the mediating effects of the LASSO-selected markers on the association between GSD history and NAFLD, a method widely applied in cross-sectional research to assess whether exposure–outcome associations may be partially accounted for by intermediate variables, thereby providing supportive and hypothesis-generating evidence for plausible causal structures. The analysis estimated direct and indirect effects via bootstrap resampling (500 iterations). All mediation models were adjusted for all covariates in Model 3.

Statistical analyses were performed using R version 4.4.1. A 2-tailed P-value <.05 was considered statistically significant. The results of logistic regression were reported as odds ratios (OR) with 95% confidence intervals (CI), and their statistical significance was tested.

3. Results

3.1. Baseline characteristics

A total of 3401 participants were included in the final analysis, of whom 324 had a history of GSD (Table 1). Compared with participants without GSD, those with GSD were significantly older (48.00 [38.00–63.25] versus 43.00 [31.00–59.00], P < .001), had higher ABSI values (0.82 [0.78–0.85] versus 0.81 [0.77–0.84], P = .001), and included a higher proportion of females (83.95% vs 62.89%, P < .001). They also reported less frequent exercise (43.52% vs 53.10%, P = .001) and higher smoking rates (41.98% vs 32.76%, P <.001). Clinically, participants with GSD had a higher prevalence of CVD (11.73% vs 6.50%, P <.001), diabetes mellitus (27.16% vs 15.18%, P <.001), hypertension (48.46% vs 34.32%, P < .001), hyperlipidemia (75.62% vs 67.01%, P = .002), lung disease (31.79% vs 17.94%, P <.001), and kidney stones (20.37% vs 7.21%, P < .001). In terms of laboratory findings, participants with GSD had lower albumin (Alb) (40.00 [37.00–42.00] vs 41.00 [39.00–43.00], P < .001), iron (13.00 [9.70–16.70] vs. 14.30 [10.40–18.60], P < .001), and Scr levels (66.30 [57.46–80.44] vs 70.72 [60.11–83.98], P < .001), but higher LDH (156.00 [135.75–177.00] vs 151.00 [134.00–170.00], P = .019) and ALP levels (77.00 [63.00–93.00] vs 71.00 [59.00–85.00], P < .001). Hematologically, participants with GSD had higher WBC (7.40 [6.18–8.72] vs 6.90 [5.60–8.30], P < .001), NEUT (4.30 [3.40–5.40] vs 3.90 [3.00–5.00], P < .001), PLT counts (256.50 [215.00–302.00] vs 247.00 [209.00–291.00], P = .012), RDW (13.80 [13.20–14.60] vs 13.50 [13.00–14.20], P < .001), and hsCRP levels (2.87 [1.29–6.33] vs 1.79 [0.78–4.30], P < .001), as well as lower RBC (4.58 [4.30–4.91] vs 4.67 [4.37–4.99], P = .004), HGB (13.40 [12.50–14.40] vs 13.80 [12.90–14.80], P < .001), and HCT (40.20 [37.60–42.70] vs 41.20 [38.50–43.70], P <.001). Most importantly, the prevalence of NAFLD was significantly higher in participants with GSD compared with those without (70.37% vs 53.85%, P < .001).

Table 1.

Baseline characteristics of the participants with and without GSD.

Characteristics	no GSD	GSD	P-value
N	3077	324	–
Age	43.00 (31.00–59.00)	48.00 (38.00–63.25)	<.001
Gender
Female	1935 (62.89%)	272 (83.95%)	<.001
Male	1142 (37.11%)	52 (16.05%)
Marital status
Married/living with partner	1822 (59.21%)	192 (59.26%)	.070
Never married	721 (23.43%)	62 (19.14%)
Widowed/divorced/separated	534 (17.35%)	70 (21.60%)
Education level
<High School	515 (16.74%)	55 (16.98%)	.800
>High School	1858 (60.38%)	190 (58.64%)
High School	704 (22.88%)	79 (24.38%)
PIR
High	732 (23.79%)	77 (23.77%)	.769
Low	508 (16.51%)	56 (17.28%)
Middle	1443 (46.90%)	156 (48.15%)
Unknown	394 (12.80%)	35 (10.80%)
Exercising status
No	1443 (46.90%)	183 (56.48%)	.001
Yes	1634 (53.10%)	141 (43.52%)
Smoking status
No	2069 (67.24%)	188 (58.02%)	<.001
Yes	1008 (32.76%)	136 (41.98%)	–
ABSI	0.81 (0.77–0.84)	0.82 (0.78–0.85)	.001
Diabetes mellitus
No	2610 (84.82%)	236 (72.84%)	<.001
Yes	467 (15.18%)	88 (27.16%)
Hypertension
No	2021 (65.68%)	167 (51.54%)	<.001
Yes	1056 (34.32%)	157 (48.46%)
Hyperlipidemia
No	1015 (32.99%)	79 (24.38%)	.002
Yes	2062 (67.01%)	245 (75.62%)
CVD
No	2877 (93.50%)	286 (88.27%)	<.001
Yes	200 (6.50%)	38 (11.73%)
Lung disease
No	2525 (82.06%)	221 (68.21%)	<.001
Yes	552 (17.94%)	103 (31.79%)
Kidney stone
No	2855 (92.79%)	258 (79.63%)	<.001
Yes	222 (7.21%)	66 (20.37%)
NAFLD
No	1420 (46.15%)	96 (29.63%)	<.001
Yes	1657 (53.85%)	228 (70.37%)
Cholecystectomy status
No	2999 (97.47%)	79 (24.38%)	<.001
Yes	78 (2.53%)	245 (75.62%)
Alb	41.00 (39.00–43.00)	40.00 (37.00–42.00)	<.001
ALP	71.00 (59.00–85.00)	77.00 (63.00–93.00)	<.001
ALT	16.00 (12.00–24.00)	17.00 (13.00–25.00)	.120
AST	18.00 (15.00–22.00)	18.00 (15.00–22.00)	.222
BAS	0.10 (0.00–0.10)	0.10 (0.00–0.10)	.016
Bicarbonate	25.00 (24.00–27.00)	25.00 (23.00–27.00)	.307
BUN	4.64 (3.93–5.71)	4.64 (3.93–6.07)	.407
Chloride	102.00 (100.00–103.00)	102.00 (100.00–104.00)	.020
EOS	0.10 (0.10–0.20)	0.20 (0.10–0.20)	.311
Folate	32.30 (22.20–47.90)	32.60 (22.57–48.40)	.291
GGT	19.00 (13.00–28.00)	19.00 (14.00–30.00)	.496
Glb	31.00 (28.00–34.00)	31.00 (28.00–34.00)	.560
HCT	41.20 (38.50–43.70)	40.20 (37.60–42.70)	<.001
HGB	13.80 (12.90–14.80)	13.40 (12.50–14.40)	<.001
hsCRP	1.79 (0.78–4.30)	2.87 (1.29–6.33)	<.001
Iron	14.30 (10.40–18.60)	13.00 (9.70–16.70)	<.001
LDH	151.00 (134.00–170.00)	156.00 (135.75–177.00)	.019
LYM	2.10 (1.70–2.60)	2.10 (1.80–2.70)	.284
MCV	88.40 (85.00–91.70)	88.10 (84.40–91.32)	.104
MON	0.50 (0.40–0.60)	0.50 (0.50–0.60)	.126
MPV	8.20 (7.60–8.80)	8.20 (7.70–8.80)	.729
NEUT	3.90 (3.00–5.00)	4.30 (3.40–5.40)	<.001
Phosphorus	1.13 (1.03–1.26)	1.16 (1.03–1.26)	.284
PLT	247.00 (209.00–291.00)	256.50 (215.00–302.00)	.012
Potassium	4.00 (3.80–4.30)	4.00 (3.80–4.30)	.688
RBC	4.67 (4.37–4.99)	4.58 (4.30–4.91)	.004
RDW	13.50 (13.00–14.20)	13.80 (13.20–14.60)	<.001
Scr	70.72 (60.11–83.98)	66.30 (57.46–80.44)	<.001
Sodium	140.00 (139.00–142.00)	141.00 (139.00–142.00)	.051
TBil	6.84 (5.13–8.55)	6.84 (5.13–8.55)	.134
UA	297.40 (243.90–356.90)	303.30 (261.70–350.90)	.160
WBC	6.90 (5.60–8.30)	7.40 (6.18–8.72)	<.001

Continuous variables were expressed as medians and IQRs, while categorical variables were expressed as percentages. For continuous variables, the P-value was based on the Wilcoxon rank-sum test, and for categorical variables, the P-value was based on the Pearson chi-square test.

ABSI = a body shape index, Alb = albumin, ALP = alkaline phosphatase, ALT = alanine transaminase, AST = aspartate aminotransferase, BAS = basophils number, BUN = blood urea nitrogen, CVD = cardiovascular disease, EOS = eosinophils number, GGT = gamma glutamyl transferase, Glb = globulin, GSD = gallstone disease, HCT = hematocrit, HGB = hemoglobin; hsCRP, high-sensitivity C-reactive protein, IQR = interquartile range, LDH = lactate dehydrogenase, LYM = lymphocyte number, MCV = mean cell volume, MON = monocyte number, MPV = mean platelet volume, NAFLD = nonalcoholic fatty liver disease, NEUT = neutrophil count, PIR = income-to-poverty ratio, PLT = platelet count, RBC = red blood cell count, RDW = red cell distribution width, Scr = serum creatinine, TBil = total bilirubin, UA = uric acid, WBC = white blood cell count.

3.2. Multivariable logistic regression for GSD-NAFLD association

In the unadjusted Model 1, GSD history significantly associated with an increased risk of NAFLD (OR = 2.035, 95% CI: 1.592–2.620, P <.001). In Model 2 (adjusted for partial covariates), the OR was 1.943 (95% CI: 1.508–2.521, P <.001). After adjusting for all predefined covariates (Model 3), GSD history remained a significant association with an increased risk of NAFLD (OR = 1.747, 95% CI: 1.334–2.302, P <.001), indicating that confounding by other covariates slightly attenuated the association but did not eliminate it (Table 2).

Table 2.

Logistic regression analysis on the association between GSD and NAFLD.

	Model 1OR (95% CI)	P-value	Model 2 OR (95% CI)	P-value	Model 3 OR (95% CI)	P-value
GSD
No	Reference	–	Reference	–	Reference	–
Yes	2.035 (1.592–2.620)	<.001	1.943 (1.508–2.521)	<.001	1.747 (1.334–2.302)	<.001

Data are presented as OR (95% CI). Model 1: unadjusted; Model 2: adjusted for age, gender, marital status, education level, and PIR; Model 3: Model 2 adjusted for exercising status, smoking status, ABSI, diabetes mellitus, hypertension, hyperlipidemia, CVD, lung disease, and kidney stone.

ABSI = a body shape index, CI = confidence interval, CVD = cardiovascular disease, GSD = gallstone disease, NAFLD = nonalcoholic fatty liver disease, OR = odds ratio, PIR = income-to-poverty ratio.

3.3. Restricted cubic spline analysis

RCS analysis (adjusted for all covariates) revealed a significant association between GSD history and liver ultrasound-derived parameters (overall term P <.001; nonlinear term P = .493 >.05) (Fig. 2). Participants with a history of GSD consistently exhibited higher median CAP, suggesting more severe steatosis compared with those without GSD, and no evidence of a nonlinear relationship was observed.

Figure 2.

Association between median CAP and NAFLD by RCS analysis. The figure displays the adjusted OR (solid lines) with 95% CI (shaded areas). CAP = controlled attenuation parameter, CI = confidence interval, NAFLD = nonalcoholic fatty liver disease, OR = odds ratio, RCS = restricted cubic spline.

3.4. Stratified analysis by cholecystectomy and sex

In sex-stratified analyses of the overall population, the association between GSD and NAFLD was not statistically significant in either men (OR = 0.84, 95% CI: 0.36–1.95, P = .681) or women (OR = 1.39, 95% CI: 0.90–2.15, P = .133). No significant interaction between sex and GSD was observed (P for interaction = 0.311) (Fig. S4, Supplemental Digital Content, https://links.lww.com/MD/R402).

Among the 324 participants with a history of GSD, 245 (75.62%) had undergone cholecystectomy (Table S1, Supplemental Digital Content, https://links.lww.com/MD/R402). In multivariable logistic regression models adjusted for all covariates, cholecystectomy status was not significantly associated with NAFLD risk in the GSD subgroup (OR = 1.537, 95% CI: 0.803–2.932, P = .192) (Table 3). Sex-stratified analyses within this subgroup similarly showed no statistically significant association between cholecystectomy and NAFLD in either men (OR = 6.05, 95% CI: 0.55–65.96, P = .14) or women (OR = 1.60, 95% CI: 0.76–3.36, P = .217), and no significant interaction between sex and cholecystectomy was detected (P for interaction = .505) (Fig. S4, Supplemental Digital Content, https://links.lww.com/MD/R402).

Table 3.

Logistic regression analysis on the association between cholecystectomy and NAFLD in participants with GSD.

	Model 1 OR (95% CI)	P-value	Model 2 OR (95% CI)	P-value	Model 3 OR (95% CI)	P-value
Cholecystectomy
No	Reference		Reference		Reference
Yes	1.324 (0.764–2.265)	.310	1.349 (0.751–2.397)	.310	1.537 (0.803–2.932)	.192

Data are presented as OR (95% CI). Model 1: unadjusted; Model 2: adjusted for age, gender, marital status, education level, and PIR; Model 3: Model 2 adjusted for exercising status, smoking status, ABSI, diabetes mellitus, hypertension, hyperlipidemia, CVD, lung disease, and kidney stone.

ABSI = a body shape index, CI = confidence interval, CVD = cardiovascular disease, GSD = gallstone disease, NAFLD = nonalcoholic fatty liver disease, OR = odds ratio, PIR = income-to-poverty ratio.

3.5. LASSO regression and mediation analysis

Because λ_min retained too many markers, making mediation analysis computationally cumbersome and potentially increasing the risk of false positives, λ_1se was adopted to select a more parsimonious set of potential mediators (Fig. S3, Supplemental Digital Content, https://links.lww.com/MD/R402, Fig. 3). Using λ_1se, LASSO regression identified 17 potential mediators: Alb, ALT, ALP, aspartate aminotransferase, BAS, HCO3, HGB, hsCRP, LDH, LYM, P, RBC, RDW, Scr, total bilirubin, UA, and WBC. Mediation analysis adjusted for all covariates showed that 4 markers (ALP, ALT, UA, and WBC) had significant mediating effects on the GSD-NAFLD association, accounting for 6.08% (P = .016), 19.54% (P <.001), 11.37% (P = .004), and 5.52% (P = .008) of the effect, respectively. (Fig. 4, Table 4).

Figure 3.

The LASSO regression analysis for identifying key NAFLD-related hematological and biochemical markers. (A) A 10-fold cross-validation of the LASSO regression model. (B) The coefficient shrinkage process of all 32 hematological and biochemical markers, changes in coefficients of different features under various levels of shrinkage wew represented by drawing lines of different colors. λ, lambda, LASSO = least absolute shrinkage and selection operator, NAFLD = nonalcoholic fatty liver disease.

Figure 4.

Schematic representation of the mediation model. GSD was modeled as the exposure, NAFLD as the outcome, and hematological or biochemical markers as mediators. Path c: total effects; Path c’: average direct effects; Path a × b: indirect effects. GSD = gallstone disease, NAFLD = nonalcoholic fatty liver disease.

Table 4.

Adjusted mediating effect of hematological and biochemical markers on the association between the GSD and NAFLD.

Marker	Total effect	Indirect effect	Direct effect	Mediation proportion	P-value
Alb	0.114334	0.006028	0.108306	5.27%	.072
ALP	0.113884	0.006926	0.106958	6.08%	.016
ALT	0.115120	0.022492	0.092628	19.54%	<.001
AST	0.114456	0.001304	0.113151	1.14%	.476
BAS	0.114678	0.000979	0.113699	0.85%	.468
Bicarbonate	0.116018	0.000155	0.115863	0.13%	.936
HGB	0.114982	0.001638	0.113344	1.42%	.464
hsCRP	0.111656	0.007149	0.104507	6.40%	.092
LDH	0.115128	0.001297	0.113832	1.13%	.356
LYM	0.113377	0.002503	0.110874	2.21%	.480
Phosphorus	0.114586	−0.000778	0.115365	−0.68%	.564
RBC	0.113754	0.004343	0.109411	3.82%	.268
RDW	0.114705	0.000222	0.114482	0.19%	.868
Scr	0.114964	0.000324	0.114640	0.28%	.784
TBil	0.114303	−0.001808	0.116111	−1.58%	.276
UA	0.112424	0.012785	0.099639	11.37%	.004
WBC	0.114342	0.006309	0.108034	5.52%	.008

Alb = albumin, ALP = alkaline phosphatase, ALT = alanine transaminase, AST = aspartate aminotransferase, BAS = basophils number, GSD = gallstone disease, HGB = hemoglobin; hsCRP, high-sensitivity C-reactive protein, LDH = lactate dehydrogenase, LYM = lymphocyte number, NAFLD = nonalcoholic fatty liver disease, RBC = red blood cell count, RDW = red cell distribution width, Scr = serum creatinine, TBil = total bilirubin, UA = uric acid, WBC = white blood cell count.

4. Discussion

This study aimed to advance understanding of the biological mechanisms linking GSD and NAFLD and to provide insights into the development of targeted strategies for their prevention and management.

The baseline comparison of participants revealed several important differences between those with and without GSD. Participants with GSD were older, more likely to be female, and had a higher prevalence of multiple comorbidities, including diabetes, hypertension, hyperlipidemia, and CVD. These findings are consistent with the established epidemiology of GSD, which is strongly influenced by female sex hormones, advancing age, and metabolic dysregulation.^[22] Moreover, patients with GSD demonstrated distinct biochemical and hematological alterations, including lower Alb levels, higher Scr and ALP, as well as pro-inflammatory profiles characterized by elevated WBC, NEUT, hsCRP, RDW, and PLT, accompanied by reduced HGB and HCT. These alterations reflect systemic inflammation and metabolic dysregulation that may predispose individuals to hepatic steatosis.^[23] Importantly, the higher prevalence of NAFLD among GSD participants (18.21% vs 9.20%) reinforces the hypothesis of shared metabolic mechanisms and the potential for bidirectional causality between the 2 diseases.

Our multivariable logistic regression analyses demonstrated a robust association between GSD and NAFLD. In the unadjusted model, participants with GSD had more than a 2-fold increased risk of NAFLD. Although the magnitude of this association was attenuated after adjusting for sociodemographic, anthropometric, lifestyle, and comorbidity factors, it remained statistically significant, with an adjusted OR of 1.747. A meta-analysis by Fan et al reported a pooled OR of 1.48 (95% CI: 1.32–1.65) for the GSD-NAFLD association, which is slightly lower than our fully adjusted estimate, likely due to differences in study populations and covariate adjustment.^[11] The persistence of this association indicates that the relationship between GSD and NAFLD cannot be explained solely by shared risk factors such as obesity, diabetes, or hypertension, and suggests that GSD itself or its underlying pathophysiological processes may directly predispose individuals to hepatic steatosis.

The RCS analysis further supported this relationship by demonstrating an association between GSD history and increased liver stiffness, as measured by transient elastography. This finding is noteworthy because the CAP not only serves as a validated indicator of hepatic steatosis but, when combined with liver stiffness measurement, can also be used to assess hepatic fibrosis, suggesting that GSD may be associated with more advanced liver injury.^[24] The nonlinear term did not reach statistical significance, indicating that the association between GSD and liver pathology was approximately linear.

Our stratified analyses showed that cholecystectomy was not significantly associated with NAFLD risk among participants with GSD. Although previous studies have reported conflicting results – suggesting that cholecystectomy may increase NAFLD risk by disrupting bile acid metabolism or that NAFLD is linked to cholecystectomy rather than gallstones themselves.^[25,26] In contrast, our findings indicate that gallbladder removal does not substantially modify the GSD–NAFLD association. By restricting the analysis to individuals with GSD, we accounted for underlying GSD-related metabolic disturbances, which may be the primary contributors to NAFLD risk. It is also possible that cholecystectomy has counterbalancing effects, simultaneously alleviating gallbladder stasis while altering bile acid circulation, resulting in an overall neutral impact on NAFLD risk.^[27] However, the cross-sectional design limits assessment of long-term postoperative effects, and prospective studies with detailed bile acid profiling are needed. Furthermore, sex-stratified analyses in both the overall population and the GSD subgroup showed no statistically significant associations between GSD or cholecystectomy and NAFLD in either men or women, with no evidence of effect modification by sex, suggesting that these associations are broadly consistent across sexes despite known sex differences in GSD epidemiology.

Perhaps the most novel contribution of our study lies in the mediation analysis. Using a rigorous LASSO regression to reduce dimensionality and avoid overfitting, we identified seventeen potential mediators, of which 4 (ALP, ALT, UA, and WBC) significantly mediated the GSD-NAFLD association. Each of these markers has plausible biological relevance. ALT, a direct marker of hepatocellular injury, is strongly correlated with hepatic fat content and necroinflammation, suggesting that part of the effect of GSD on NAFLD may be mediated through hepatocellular stress. ALP, an enzyme involved in bile acid synthesis, when elevated, often indicates cholestasis or biliary tract dysfunction and may reflect the overlap between gallstone-related hepatic inflammation, impaired bile flow, and hepatic lipid accumulation.^[28] Elevated UA has been consistently linked to metabolic syndrome, oxidative stress, and NAFLD risk, supporting its role as both a marker and a mediator of liver fat accumulation.^[29] Similarly, WBC, a nonspecific but sensitive indicator of systemic inflammation, may capture the contribution of chronic low-grade inflammation and immune activation in linking GSD with NAFLD, given the established role of inflammatory pathways in hepatic steatosis and progression to nonalcoholic steatohepatitis.^[30] The relatively small mediation proportions suggest that these markers explain only part of the GSD-NAFLD association, indicating that other unmeasured mediators may play more prominent roles. Moreover, these mediators may act synergistically, and future studies should investigate potential interactive effects among them.

The association between GSD and NAFLD can be attributed to several overlapping pathophysiological mechanisms. GSD is characterized by bile cholesterol supersaturation, impaired bile acid homeostasis, and gallbladder dysmotility, all of which may disrupt enterohepatic bile acid signaling, alter lipid metabolism, and induce systemic oxidative stress and inflammation, thereby facilitating the development of NAFLD.^[31] Dysregulation of bile acid metabolism represents a central pathway: increased bile cholesterol saturation promotes gallstone formation, while impaired enterohepatic bile acid circulation compromises hepatic bile acid synthesis and transport. This in turn affects key bile acid receptors such as FXR and TGR5, undermining their regulatory roles in glucose, lipid, and energy metabolism, and thereby predisposing to metabolic dysfunction-associated steatotic liver disease and metabolic syndrome.^[27] In parallel, chronic low-grade inflammation in GSD, characterized by elevated WBC, CRP, and TNF-α, may exacerbate hepatocellular injury, as reflected by increased ALP and ALT levels, while elevated UA promotes oxidative stress and mitochondrial dysfunction in hepatocytes.^[32,33] insulin resistance, common to both GSD and NAFLD, plays a pivotal role by reducing gallbladder motility, which predisposes to gallstone formation, and simultaneously promoting hepatic lipogenesis, thereby accelerating steatosis.^[34,35] Together, these interrelated mechanisms underscore the multifaceted biological crosstalk between gallbladder disease and liver metabolism and provide a plausible explanation for the independent association observed between GSD history and NAFLD.

This study has several notable strengths. By leveraging a large, nationally representative NHANES sample, our findings are broadly generalizable. We conducted comprehensive covariate adjustments, thereby minimizing residual confounding. Methodologically, we applied RCS analysis to assess potential nonlinear associations, performed stratified analyses among GSD patients to clarify the role of cholecystectomy and sex, and, to our knowledge, were the first to use LASSO regression for mediator selection in the GSD-NAFLD context, thus reducing bias from arbitrary mediator choice. Moreover, we employed formal mediation analysis to quantify mediating effects, providing novel mechanistic insights.

Despite its strengths, this study has several limitations. First, its cross-sectional design precludes causal inference. Although our findings suggest a potential causal pathway from GSD to NAFLD, reverse causality cannot be excluded, as hepatic steatosis may alter bile composition and increase the likelihood of gallstone formation.^[11] Prospective longitudinal studies are therefore needed to establish the temporal sequence of these associations. Second, GSD history was based on self-report, which may introduce recall bias and misclassification. Nevertheless, because gallstone diagnosis typically involves imaging and medical consultation, self-reported history in this context is likely reasonably reliable. Third, although we adjusted for a wide range of covariates, residual confounding from unmeasured factors such as dietary intake, gut microbiome composition, or detailed hormonal status cannot be ruled out. Finally, although mediation analysis identified significant biochemical mediators, the presence of unmeasured parallel pathways remains possible, and the identified mediators should be interpreted as partial contributors rather than exclusive mechanisms.

Our findings underscore the importance of considering GSD in NAFLD risk stratification and of monitoring patients even after cholecystectomy. Although ALP, ALT, UA, and WBC were identified as mediators, their modest mediating effects suggest that no single pathway fully accounts for the observed association, and that effective prevention or intervention strategies may need to target multiple metabolic processes. Importantly, the fact that many of these markers remain within reference ranges does not diminish their relevance in population-based research, where subclinical shifts and distributional differences often capture early or chronic metabolic dysregulation. Future longitudinal and interventional studies are warranted to clarify causality, investigate additional mediators such as bile acid metabolites, cytokines, and adipokines, and evaluate the roles of genetic and microbiome factors in shaping the GSD-NAFLD relationship.^[36]

5. Conclusions

The present study demonstrates that GSD is independently associated with an increased risk of NAFLD in adults, and this relationship is not modified by cholecystectomy. The association appears to be partially mediated by biochemical markers such as ALP, ALT, UA, and WBC. These findings highlight the complex interplay between biliary and hepatic disorders and emphasize the need for integrated management strategies in patients with GSD who may be at higher risk for NAFLD. Future longitudinal and mechanistic studies are warranted to confirm causality, elucidate underlying biological pathways, and determine whether interventions targeting these mediators can reduce NAFLD risk in this population.

Acknowledgments

We would like to thank all participants in this study.

Author contributions

Conceptualization: Rongxuan Li, Bingchen Wang, Xiao Chen.

Data curation: Rongxuan Li.

Formal analysis: Rongxuan Li, Bingchen Wang, Xiao Chen.

Investigation: Rongxuan Li, Bingchen Wang, Xiao Chen.

Methodology: Rongxuan Li, Xiao Chen.

Software: Rongxuan Li, Bingchen Wang, Angshu Cai.

Supervision: Dayong Cao, Jianguo Zhou.

Visualization: Rongxuan Li.

Writing – original draft: Rongxuan Li.

Writing – review & editing: Rongxuan Li, Bingchen Wang, Xiao Chen.

Supplementary Material

medi-105-e47759-s001.pdf ^{(882.9KB, pdf)}