Small dense low-density lipoprotein in primary biliary cholangitis: Independent association with metabolic syndrome

Abstract

BACKGROUND

Primary biliary cholangitis (PBC) is frequently associated with cholestasis-driven hypercholesterolemia characterized by lipoprotein X accumulation. Despite markedly elevated cholesterol levels, cardiovascular risk in PBC is not consistently increased, suggesting that conventional lipid parameters may inadequately reflect true atherogenic burden. Small dense low-density lipoprotein (sdLDL), a highly atherogenic low-density lipoprotein (LDL) subfraction closely linked to insulin resistance and metabolic syndrome (MetS), has not been systematically evaluated in PBC.

AIM

To evaluate determinants of sdLDL levels in PBC, focusing on MetS, metabolic dysfunction-associated steatotic liver disease (MASLD), liver fibrosis, and biochemical response, and assess whether sdLDL reflects metabolic dysfunction rather than cholestatic disease severity.

METHODS

In this cross-sectional study, 148 consecutive patients with PBC were recruited from a tertiary hepatology center. sdLDL was quantified using the Lipoprint^® LDL System. MetS was defined according to International Diabetes Federation criteria. Hepatic steatosis and fibrosis were assessed using vibration-controlled transient elastography. Multivariable linear and logistic regression analyses were performed to evaluate independent determinants of sdLDL.

RESULTS

MetS was present in 39.2% of patients. sdLDL levels were significantly higher in patients with MetS (P = 0.004). In multivariable analysis, MetS remained independently associated with ln-transformed sdLDL levels [B = 0.76; 95% confidence interval (CI): 0.26-1.26; P = 0.003] and with the presence of detectable sdLDL (odds ratio 2.10; 95%CI: 1.02-4.35; P = 0.045), independent of age, sex, adiposity, and lipid-lowering therapy. In contrast, MASLD, advanced fibrosis, and cholestatic activity were not independently associated with sdLDL. Notably, LDL-C levels were lower in patients with MetS, whereas sdLDL and sdLDL/largeLDL ratio were significantly increased, suggesting discordance between conventional LDL-C and qualitative lipoprotein remodeling.

CONCLUSION

In PBC, sdLDL appears to reflect superimposed metabolic dysfunction rather than cholestatic liver disease severity. Advanced lipoprotein profiling may therefore help distinguish metabolically driven atherogenic dyslipidemia from cholestatic hypercholesterolemia in contemporary PBC cohorts.

Key Words: Primary biliary cholangitis; Small dense low-density lipoprotein; Metabolic dysfunction; Lipoprotein subfractions; Atherogenic dyslipidemia

Core Tip: Hypercholesterolemia in primary biliary cholangitis (PBC) is largely driven by cholestasis and lipoprotein X, making conventional lipid parameters difficult to interpret. In this study, small dense low-density lipoprotein (LDL) was independently associated with metabolic syndrome but not with liver fibrosis, cholestatic activity, metabolic dysfunction-associated steatotic liver disease, or biochemical response. These findings suggest that small dense LDL reflects superimposed metabolic dysfunction rather than intrinsic cholestatic liver disease and may help identify a metabolically driven atherogenic phenotype in contemporary PBC cohorts.

INTRODUCTION

Primary biliary cholangitis (PBC) is a chronic autoimmune cholestatic liver disease characterized by progressive destruction of small intrahepatic bile ducts, leading to cholestasis and, in a subset of patients, progression to cirrhosis[1,2]. It predominantly affects women and has a stable incidence but increasing prevalence in developed countries[3]. Diagnosis is based on persistent cholestatic liver enzyme elevation, PBC-specific autoantibodies, and compatible histology[1].

Recent guidelines on PBC recommend addressing and treating several extrahepatic diseases, including osteoporosis, celiac disease, thyroid disease, and Sjögren’s syndrome. However, only these specific extrahepatic diseases are mentioned, and the level of evidence is low to moderate. Furthermore, the prevalence of these extrahepatic diseases varied significantly across studies, making it difficult to assess the extent and severity of these conditions[1,4].

Ursodeoxycholic acid (UDCA) remains first-line therapy, while second-line treatment options are available for patients with incomplete biochemical response[5,6]. Response to UDCA therapy is commonly evaluated using biochemical parameters. Complete biochemical response (CBR) was defined as normalization of serum alkaline phosphatase (ALP) and bilirubin concentrations, although other validated response criteria, such as the Toronto criteria [ALP < 1.67 × upper limit of normal (ULN)], are also used in clinical practice[7]. Biochemical response is associated with improved prognosis[8].

Cholestasis profoundly alters lipid metabolism in PBC. Impaired bile acid secretion promotes the formation of lipoprotein X (LpX), an abnormal phospholipid-rich particle lacking apolipoprotein B (ApoB)[9,10]. Impaired bile acid secretion causes phospholipids and free cholesterol to accumulate in the liver and subsequently enter the circulation. These lipids then associate with albumin to form LpX, an unusual ApoB-deficient lipoprotein that remains in circulation since it is not removed via low-density lipoprotein (LDL) receptors. As a result, standard LDL-C measurements can overstate the actual amount of atherogenic lipoproteins in patients with PBC[9]. Hypercholesterolemia in PBC differs from classic metabolic dyslipidemia in that it is primarily caused by the accumulation of LpX. In contrast, metabolic dyslipidemia is characterized by elevated concentrations of ApoB-containing lipoproteins, hypertriglyceridemia, insulin resistance, and an increased proportion of small, dense LDL particles, which contribute significantly to the development of atherosclerosis[9,11,12]. Consequently, hypercholesterolemia occurs in up to 75%-95% of patients[13,14]. Because LpX overlaps with LDL in density, routine lipid panels may report elevated LDL-C levels, potentially mimicking a proatherogenic lipid profile[9].

Despite markedly elevated total cholesterol levels, cardiovascular risk in PBC is not consistently increased[13,15]. Experimental data suggest that LpX-containing fractions are less susceptible to oxidative modification and may attenuate LDL atherogenicity[16]. Thus, hypercholesterolemia in PBC appears qualitatively distinct from classical metabolic dyslipidemia. Importantly, conventional LDL-C measurements may inadequately reflect true atherogenic burden and qualitative lipoprotein remodeling in cholestatic PBC. However, detailed characterization of lipoprotein subfractions in this setting remains limited.

Small dense LDL (sdLDL) is a particularly atherogenic LDL subfraction. Compared with larger LDL particles, sdLDL exhibits prolonged circulation time, increased arterial wall penetration, and greater susceptibility to oxidative modification[12,17]. Elevated sdLDL levels are closely linked to insulin resistance and hypertriglyceridemia and are commonly observed in metabolic syndrome (MetS) and type 2 diabetes[11]. In prospective cohorts, sdLDL-C predicts incident coronary heart disease independently of traditional lipid parameters[11,12,18]. Whether sdLDL in PBC reflects metabolically driven atherogenic dyslipidemia or cholestatic lipid remodeling remains unknown.

The clinical phenotype of PBC is evolving. Contemporary cohorts increasingly include patients with metabolic dysfunction and hepatic steatosis. Metabolic dysfunction-associated steatotic liver disease (MASLD), previously classified as metabolic associated fatty liver disease, has been reported across autoimmune liver diseases, including PBC[19,20], while coexisting non-alcoholic fatty liver disease has been described in up to 24% of patients[21]. Nearly half of patients in observational cohorts fulfill criteria for MASLD[22]. In this context, atherogenic lipoprotein patterns may increasingly reflect superimposed metabolic dysfunction rather than cholestasis itself.

To date, determinants of sdLDL levels in PBC have not been systematically evaluated. It remains unclear whether sdLDL primarily reflects cholestatic liver disease severity or an underlying metabolic phenotype independent of hepatic factors. Clarifying this distinction may improve interpretation of dyslipidemia and refine cardiovascular risk assessment in PBC. Although the association between sdLDL-C and hepatic steatosis has recently been described in patients with MASLD[23], little is still known about the clinical significance of sdLDL in cholestatic liver diseases such as PBC, particularly given the distinct nature of lipid metabolism disorders compared to metabolic dyslipidemia. Therefore, the aim of the present study was to evaluate determinants of sdLDL levels in patients with PBC, with particular focus on MetS, MASLD, fibrosis stage, and biochemical response.

MATERIALS AND METHODS

We conducted a cross-sectional observational study including consecutive patients with PBC recruited from the Hepatology Outpatient Clinic of the 2^nd Department of Internal Medicine, Louis Pasteur University Hospital and Pavol Jozef Šafárik University Faculty of Medicine in Košice, Slovakia. Patients were consecutively enrolled during a predetermined recruitment period from January to June 2024. Statistical analyses were performed after completion of patient recruitment and database verification.

All participants fulfilled diagnostic criteria for PBC according to the European Association for the Study of the Liver (EASL) guidelines[1]. Exclusion criteria included refusal to participate and presence of alternative causes of hepatic steatosis or chronic liver disease. All patients underwent comprehensive etiological evaluation, including testing for viral hepatitis B and C and assessment for other liver diseases.

Study visits included laboratory testing, vibration-controlled transient elastography (VCTE), anthropometric assessment, and structured clinical data collection. Use of UDCA, fibrates, and statins at the time of lipoprotein assessment was recorded. Information on concomitant autoimmune diseases was not collected or evaluated as part of the study protocol, therefore, it has not been included in further analyses. The study protocol was approved by the Ethics Committee of Louis Pasteur University Hospital in Košice (reference No. 2020/EK10076). Patient recruitment was delayed due to the coronavirus disease 2019 pandemic.

Definitions

MetS was diagnosed according to the International Diabetes Federation criteria[24].

Hepatic steatosis was assessed using VCTE with controlled attenuation parameter (CAP). A CAP value ≥ 248 dB/m defined hepatic steatosis, while thresholds of 248 dB/m, 268 dB/m, and 280 dB/m corresponded to steatosis grades ≥ S1, ≥ S2, and S3 according to current EASL-EASD-EASO clinical practice guidelines on the management of MASLD, respectively[25].

MASLD was defined as hepatic steatosis in the presence of at least one cardiometabolic risk factor and absence of alternative causes of steatotic liver disease[25].

Liver fibrosis was categorized according to VCTE-derived liver stiffness as mild-to-moderate fibrosis (F0-F2; < 10.7 kPa) or advanced fibrosis (F3/F4; ≥ 10.7 kPa)[26,27].

CBR was defined as normalization of ALP and total bilirubin levels.

Lipoprint subfractions analysis

Lipoprotein subfractions were analyzed using the Lipoprint^® LDL System (Quantimetrix Corporation, Redondo Beach, CA, United States), a linear polyacrylamide gel electrophoresis method separating lipoproteins according to particle size.

Fasting venous blood samples were obtained after ≥ 12 hours of fasting. Serum samples were processed according to manufacturer recommendations. Briefly, 25 µL of serum was mixed with 200 µL of loading gel containing lipophilic dye, photopolymerized, and subjected to electrophoresis at 3 mA per gel tube (maximum 500 V) for approximately 60 minutes.

Fractions were separated into very low-density lipoprotein (VLDL), intermediate-density lipoproteins (midbands A-C), LDL-1 to LDL-7, and high-density lipoprotein (HDL) based on relative electrophoretic mobility. Bands were scanned and quantified using Lipoware software. Cholesterol content of each subfraction was calculated from relative band area and total cholesterol concentration.

sdLDL was defined as the sum of LDL-3 through LDL-7 subfractions[28].

Exposure and outcome definitions

The primary exposure of interest was MetS, defined according to the above-mentioned established clinical criteria. Additional exposure variables included the presence of MASLD, advanced liver fibrosis, CBR, and lipid-lowering therapy. Current use of statins and fibrates at the time of lipoprotein subfraction assessment was recorded and included as covariates in multivariable analyses.

The primary outcome was serum sdLDL concentration. Ln-transformed sdLDL was used as the continuous dependent variable in linear regression analyses. As a secondary outcome, detectable sdLDL (sdLDL > 0) was analyzed as a binary variable using logistic regression models to explore its association with metabolic and liver-related factors.

Additional exploratory analyses included assessment of conventional and advanced lipoprotein-related parameters, including LDL-C, ApoB, lipoprotein(a) [Lp(a)], ApoB/apolipoprotein A-I (ApoA-I) ratio, large LDL subfractions, and sdLDL/LargeLDL ratio. Receiver operating characteristic (ROC) analyses were performed to evaluate the discriminatory performance of sdLDL, LDL-C, large LDL, and sdLDL/LargeLDL ratio for identification of MetS.

Statistical analysis

Continuous variables were expressed as mean ± SD or median with interquartile range (IQR), as appropriate, based on data distribution. Normally distributed variables were analyzed using independent-samples t-test and presented as mean ± SD, whereas non-normally distributed variables were analyzed using Mann-Whitney U test and presented as median (IQR). Categorical variables were presented as absolute numbers and percentages. Between-group comparisons were performed using the independent-samples t-test or Mann-Whitney U test, as appropriate. No formal sample size calculation was performed because this was an exploratory observational study.

The primary multivariable analyses were prespecified to evaluate the association between MetS and sdLDL, given the central role of MetS in atherogenic dyslipidemia.

Because sdLDL values demonstrated a right-skewed distribution and included zero values, a small constant (0.01 mmol/L) was added before logarithmic transformation. Ln-transformed sdLDL was calculated as ln (sdLDL + 0.01) and used in linear regression analyses. In addition, a substantial proportion of participants had undetectable sdLDL concentrations (sdLDL = 0), resulting in a partially zero-inflated distribution. Therefore, a complementary logistic regression analysis using detectable sdLDL (sdLDL > 0) as a binary outcome was performed to assess the robustness of the observed associations. Multivariable linear regression models were constructed using ln-transformed sdLDL as the dependent variable. The primary model included MetS, age, sex, and body mass index (BMI) as covariates, while a second model additionally adjusted for statin and fibrate therapy. Since MetS was included in the multivariate regression models as a binary variable defined according to standard diagnostic criteria, its individual components (waist circumference, blood pressure, triglycerides, HDL-C, and fasting glucose) were not included in the models separately, in order to avoid multicollinearity. To assess robustness of the findings, a sensitivity analysis was performed replacing BMI with waist circumference.

Additional exploratory analyses evaluated associations between MetS and lipoprotein-related parameters, including LDL-C, ApoB, Lp(a), large LDL subfractions, and sdLDL/LargeLDL ratio. ROC analyses were performed as exploratory analyses to compare the ability of lipoprotein-related parameters to identify the metabolic phenotype.

Multicollinearity was assessed using variance inflation factors (VIF) and condition index. Logistic regression analyses were performed to evaluate factors associated with detectable sdLDL (sdLDL > 0). Odds ratio (OR) with 95% confidence intervals (CI) were calculated. Model calibration was assessed using the Hosmer-Lemeshow goodness-of-fit test. Additional lipoprotein-related comparisons and ROC analyses were considered exploratory and were therefore interpreted cautiously without formal correction for multiple testing.

All statistical analyses were conducted using SPSS software (IBM Corp., Armonk, NY, United States). A two-sided P value < 0.05 was considered statistically significant.

RESULTS

Baseline characteristics

A total of 148 patients with PBC were included in the analysis. The cohort was predominantly female (92.6%), with a mean age of 64.2 ± 10.7 years.

MetS was present in 39.2% of patients, while MASLD was identified in 42.6%. Advanced fibrosis (F3/F4 by transient elastography) was observed in 21.6% of the cohort, and CBR was achieved in 55.4%.

At the time of lipoprotein assessment, 19.6% of patients were receiving statins and 23.0% fibrates. The mean sdLDL concentration in the overall cohort was 0.09 ± 0.17 mmol/L. Because sdLDL values demonstrated a right-skewed distribution and included zero values, ln-transformed sdLDL values calculated as ln (sdLDL + 0.01) were used in subsequent regression analyses. Detailed demographic, clinical, biochemical, and lipoprotein-related characteristics are summarized in Table 1.

Table 1 Baseline characteristics of the study population, n (%).

Number of patients	n = 148
Demographic and anthropometric characteristics
Age (years)	64.15 ± 10.73 (33-88)
BMI	26.76 ± 4.47 (16.51-40.16)
Waist circumference (cm)	89.36 ± 11.27 (60-121)
Gender	Female n = 137 (92.60); Male n = 11 (7.40)
Cardiometabolic comorbidities
Coronary artery disease	4 (2.70)
Acute coronary syndrome	3 (2.00)
History of stroke	7 (4.70)
Peripheral artery disease	2 (1.40)
Treatment
Statin therapy	29 (19.60)
Fibrate therapy	34 (23.00)
UDCA therapy	148 (100.00)
Liver disease characteristics
Complete biochemical response	82 (55.40)
Advanced liver fibrosis F3/F4	32 (21.60)
Metabolic syndrome	58 (39.20)
MASLD	63 (42.60)
CAP (dB/m)	244.87 ± 58.07
TE (kPa)	8.92 ± 8.16
AST × ULN (%)	101.52 ± 60.85
ALT × ULN (%)	87.43 ± 65.42
GGT × ULN (%)	212.10 ± 372.79
ALP × ULN (%)	109.12 ± 84.10
Total bilirubin (µmol/L)	13.71 ± 7.75
Conjugated bilirubin (µmol/L)	3.20 ± 2.75
Metabolic and lipoprotein parameters
Glucose (mmol/L)	5.83 ± 1.87
C-peptide (ng/mL)	2.08 ± 1.2
HOMA-IR	6.37 ± 4.47
Total cholesterol (mmol/L)	5.32 ± 1.26
HDL-C (mmol/L)	1.67 ± 0.44
LDL-C (mmol/L)	3.27 ± 0.92
Triglycerides (mmol/L)	1.30 ± 0.89
Platelets × 109/L	245.29 ± 97.35
sdLDL (mmol/L)	0.09 ± 0.17
largeLDL (mmol/L)	1.29 ± 0.49
sdLDL/LargeLDL	0.08 ± 0.18
ApoB (mg/dL)	43.91 ± 16.87
Lp(a) (mg/dL)	10.75 ± 15.34

ALP: Alkaline phosphatase; ALT: Alanine aminotransferase; ApoB: Apolipoprotein B; AST: Aspartate aminotransferase; BMI: Body mass index; CAP: Controlled attenuation parameter; F3/F4: Advanced liver fibrosis stages 3 and 4; GGT: Gamma-glutamyl transferase; HDL: High-density lipoprotein; HOMA-IR: Homeostasis model assessment of insulin resistance; LDL: Low-density lipoprotein; Lp(a): Lipoprotein a; MASLD: Metabolic dysfunction-associated steatotic liver disease; sdLDL: Small dense low-density lipoprotein; TE: Transient elastography; ULN: Upper limit of normal; UDCA: Ursodeoxycholic acid.

Univariable analyses

In univariable analyses, sdLDL levels were significantly higher in patients with MetS compared with those without MetS [median 0.07 (IQR 0.00-0.14) vs 0.00 (IQR 0.00-0.07); P = 0.004] (Figure 1).

Figure 1 Distribution of ln-transformed small dense low-density lipoprotein levels according to metabolic syndrome status. Ln (small dense low-density lipoprotein) levels were significantly higher in patients with metabolic syndrome than in those without metabolic syndrome (P = 0.004, Mann-Whitney U test). Boxplots represent median values, interquartile ranges, and whiskers. sdLDL: Small dense low-density lipoprotein.

In contrast, no significant differences in sdLDL concentrations were observed according to MASLD status (P = 0.324) or presence of advanced fibrosis (P = 0.155), suggesting that sdLDL levels were more closely associated with metabolic phenotype than liver disease severity.

Patients with CBR exhibited higher sdLDL levels compared with patients without CBR [median 0.06 (IQR 0.00-0.13) vs 0.00 (IQR 0.00-0.07); P = 0.017].

Univariable comparisons of sdLDL according to metabolic and liver-related factors are summarized in Table 2.

Table 2 Univariable comparison of small dense low-density lipoprotein levels according to metabolic and liver-related factors.

Variable	Group	sdLDL (mmol/L), median (IQR)	P value
MetS	Present	0.07 (0.00-0.14)	0.004
	Absent	0.00 (0.00-0.07)
MASLD	Present	0.03 (0.00-0.13)	0.324
	Absent	0.03 (0.00-0.08)
Advanced fibrosis (F3/F4)	Present	0.00 (0.00-0.08)	0.155
	Absent	0.04 (0.00-0.12)
CBR	Present	0.06 (0.00-0.13)	0.017
	Absent	0.00 (0.00-0.07)

CBR: Complete biochemical response; F3/F4: Fibrosis stages 3 and 4; MASLD: Metabolic dysfunction-associated steatotic liver disease; sdLDL: Small dense low-density lipoprotein; IQR: Interquartile range; MetS: Metabolic syndrome.

Multivariable linear regression analyses

In multivariable linear regression analyses using ln-transformed sdLDL as the dependent variable, MetS was independently associated with higher ln(sdLDL) levels (B = 0.72, 95%CI: 0.23-1.21; standardized β = 0.26; P = 0.005), whereas age, sex, and BMI were not significantly associated with ln(sdLDL).

After additional adjustment for lipid-lowering therapy, the association between MetS and ln(sdLDL) remained statistically significant (B = 0.76, 95%CI: 0.26-1.26; standardized β = 0.27; P = 0.003). Neither statin nor fibrate therapy was independently associated with ln(sdLDL). Although fibrates may influence LDL particle remodeling through triglyceride reduction and PPAR-α activation, fibrate therapy was not independently associated with sdLDL in our cohort, possibly reflecting the cross-sectional design and confounding by indication. The fully adjusted model was statistically significant but showed modest explanatory performance, explaining 9.3% of variance in ln(sdLDL) (adjusted R² = 0.054) (Table 3 and Figure 2).

Figure 2 Multivariable linear regression analysis of factors associated with ln-transformed small dense low-density lipoprotein levels. Error bars represent 95% confidence interval. Metabolic syndrome remained independently associated with higher ln (small dense low-density lipoprotein) levels after adjustment for age, sex, body mass index, and lipid-lowering therapy. BMI: Body mass index; CI: Confidence interval.

Table 3 Multivariable linear regression analysis of factors associated with ln-transformed small dense low-density lipoprotein levels.

Variable	B (95%CI)	Standardized β	P value
MetS	0.76 (0.26 to 1.26)	0.27	0.003
Age	-0.004 (-0.03 to 0.02)	-0.03	0.750
Female sex	-0.14 (-1.00 to 0.72)	-0.03	0.745
BMI	0.024 (-0.03 to 0.07)	0.08	0.354
Statin therapy	-0.45 (-1.04 to 0.13)	-0.13	0.133
Fibrate therapy	-0.01 (-0.53 to 0.51)	-0.004	0.964
Model statistics: R2 = 0.093 adjusted R2 = 0.054 F = 2.393 P = 0.031

BMI: Body mass index; CI: Confidence interval; MetS: Metabolic syndrome.

No evidence of clinically relevant multicollinearity was observed (maximum VIF 1.30; minimum tolerance 0.80).

Given the size of the cohort and the numerous zero values, we performed bootstrapping. Bootstrap resampling (1000 samples) confirmed the robustness of the regression model. The association between MetS and ln-transformed sdLDL remained statistically significant (bootstrap B = 0.760, bias = -0.009, BCa 95%CI: 0.265-1.229).

To further explore the relationship between cholestatic activity and sdLDL, an additional model including ALP as a continuous marker of cholestasis was constructed. ALP was not independently associated with ln(sdLDL) (B = -0.001, P = 0.315), whereas MetS remained a significant predictor (B = 0.716, P = 0.005). Inclusion of ALP did not materially improve model performance.

In a sensitivity analysis adjusted only for age and sex, MetS remained independently associated with ln(sdLDL) (B = 0.763, P = 0.002), suggesting that the association was not solely attributable to anthropometric overlap.

Sensitivity analysis

In a sensitivity analysis replacing BMI with waist circumference as a marker of central adiposity, MetS remained independently associated with higher ln-transformed sdLDL levels (B = 0.68, 95%CI: 0.17-1.18; standardized β = 0.24; P = 0.010). Waist circumference itself was not independently associated with ln(sdLDL) (B = 0.01, 95%CI: -0.01 to 0.03; P = 0.393). Results of the sensitivity analysis are summarized in Table 4.

Table 4 Sensitivity analysis using waist circumference instead of body mass index in multivariable linear regression.

Variable	B (95%CI)	Standardized β	P value
Metabolic syndrome	0.68 (0.17 to 1.18)	0.24	0.010
Age (years)	-0.01 (-0.03 to 0.01)	-0.07	0.421
Female sex	-0.02 (-0.83 to 0.86)	0.00	0.969
Waist circumference (cm)	0.01 (-0.01 to 0.03)	0.08	0.393

CI: Confidence interval.

Logistic regression analysis

In a multivariable logistic regression model adjusted for age, sex, and lipid-lowering therapy (statins and fibrates), MetS was independently associated with the presence of detectable sdLDL (OR 2.10, 95%CI: 1.02-4.35; P = 0.045).

Age, sex, statin therapy, and fibrate therapy were not significantly associated with detectable sdLDL. The model demonstrated acceptable calibration according to the Hosmer-Lemeshow goodness-of-fit test (χ² = 11.79, df = 8; P = 0.161).

Results of the multivariable logistic regression analysis are presented in Table 5.

Table 5 Multivariable logistic regression analysis of factors associated with detectable small dense low-density lipoprotein.

Variable	OR (95%CI)	P value
Metabolic syndrome	2.10 (1.02-4.35)	0.045
Age (years)	0.99 (0.96-1.03)	0.731
Female sex	1.43 (0.37-5.49)	0.606
Fibrate therapy	0.92 (0.41-2.04)	0.832
Statin therapy	0.51 (0.21-1.27)	0.149

Model adjusted for age, sex, statin therapy, and fibrate therapy. Hosmer-Lemeshow goodness-of-fit test: χ² = 11.79, df = 8; P = 0.161. OR: Odds ratio; CI: Confidence interval.

Lipoprotein remodeling according to MetS status

Patients with MetS exhibited a distinct atherogenic lipoprotein phenotype compared with patients without MetS (Table 6). Despite significantly lower LDL-C levels, patients with MetS demonstrated higher sdLDL concentrations, lower large LDL subfractions, and increased sdLDL/LargeLDL and triglycerides/HDL-C ratios. HDL-C levels were also significantly lower in patients with MetS.

Table 6 Lipoprotein subfraction phenotype in primary biliary cholangitis patients with and without metabolic syndrome.

Variable	No MetS	MetS	P value
LDL-C (mmol/L)	3.37 ± 0.82	3.08 ± 1.08	0.0211
HDL-C (mmol/L)	1.81 ± 0.38	1.44 ± 0.43	< 0.0011
sdLDL (mmol/L)	0.00 IQR (0.00-0.07)	0.07 IQR (0.00-0.14)	0.0042
largeLDL (mmol/L)	1.27 IQR (1.03-1.78)	1.09 IQR (0.87-1.49)	0.0412
TG (mmol/L)	1.04 IQR (0.82-1.29)	1.31 IQR (1.03-1.78)	< 0.0012
TG/HDL-C ratio	0.60 IQR (0.42-0.79)	1.02 IQR (0.61-1.47)	< 0.0012
sdLDL/LargeLDL ratio	0.00 IQR (0.00-0.05)	0.05 IQR (0.00-0.15)	< 0.0012

¹Independent-samples t-test.

²Mann-Whitney U test.

HDL: High-density lipoprotein cholesterol; LDL: Low-density lipoprotein cholesterol; MetS: Metabolic syndrome; sdLDL: Small dense low-density lipoprotein; IQR: Interquartile range; TG: Triglycerides.

This paradoxical pattern suggests that conventional LDL-C measurements may inadequately reflect qualitative atherogenic lipoprotein remodeling in PBC patients with superimposed metabolic dysfunction.

Additional lipoprotein-related parameters according to MetS status

ApoB concentrations, Lp(a), and ApoB/ApoA-I ratio did not differ significantly between patients with and without MetS (Table 7).

Table 7 Additional lipoprotein-related parameters in primary biliary cholangitis patients with and without metabolic syndrome.

Variable	No MetS	MetS	P value
Lp(a) (mg/dL)	3.76 IQR (1.56-13.99)	5.33 IQR (1.65-11.88)	0.8021
ApoB (mg/dL)	39.69 IQR (32.02-52.92)	41.61 IQR (31.60-52.16)	0.5601
ApoB/ApoA-I	0.59 IQR (0.41-0.76)	0.64 IQR (0.47-0.83)	0.0621

¹Mann Whitney U test.

MetS: Metabolic syndrome; Lp(a): Lipoprotein a; ApoA-I: Apolipoprotein A-I; IQR: Interquartile range; ApoB: Apolipoprotein B.

Despite the absence of significant differences in ApoB-related parameters, patients with MetS demonstrated a markedly more atherogenic lipoprotein subfraction profile characterized by higher sdLDL levels and increased sdLDL/LargeLDL ratio.

Discriminatory performance of lipoprotein-related parameters for MetS

ROC analyses were performed to evaluate the discriminatory performance of lipoprotein-related parameters for identification of MetS in patients with PBC (Figure 3 and Table 8). Optimal cut-off values were derived using the maximum Youden index and are presented for descriptive purposes only. Given the exploratory nature of the ROC analyses, these thresholds should not be interpreted as clinically applicable diagnostic cut-offs.

Figure 3 Receiver operating characteristic analysis of lipoprotein-related parameters for identification of metabolic syndrome. Receiver operating characteristic curves are shown for small dense low-density lipoprotein (sdLDL), low-density lipoprotein (LDL)-C, large LDL, and sdLDL/LargeLDL ratio. The sdLDL/LargeLDL ratio demonstrated the highest discriminatory performance, whereas LDL-C and large LDL showed inverse discriminatory patterns. Curves for LDL-C and large LDL fall below the diagonal reference line, indicating that lower values of these parameters were associated with metabolic syndrome. sdLDL: Small dense low-density lipoprotein; ROC: Receiver operating characteristic; MetS: Metabolic syndrome; LDL: Low-density lipoprotein; PBC: Primary biliary cholangitis.

Table 8 Receiver operating characteristic analysis of lipoprotein-related parameters for identification of metabolic syndrome.

Variable	AUC (95%CI)	P value	Optimal cut-off	Sensitivity (%)	Specificity (%)	Youden index
sdLDL	0.634 (0.540-0.729)	0.006	0.075	46.6	22.2	0.243
largeLDL	0.400 (0.305-0.495)	0.048	2.290	6.9	97.8	0.047
LDL-C	0.378 (0.283-0.473)	0.013	5.360	6.9	98.9	0.058
sdLDL/LargeLDL ratio	0.650 (0.554-0.745)	0.002	0.034	62.1	70.0	0.321

LDL: Low-density lipoprotein cholesterol; sdLDL: Small dense low-density lipoprotein; AUC: Area under the curve; CI: Confidence interval.

Among the evaluated parameters, the sdLDL/LargeLDL ratio showed the highest discriminatory ability, although overall discrimination remained modest, whereas lower LDL-C and lower large LDL levels were associated with the presence of MetS. This inverse association is reflected by ROC curves falling below the diagonal reference line. Overall, sdLDL-related parameters showed modestly better discriminatory performance than conventional LDL-C measurements for identification of the metabolically remodeled lipoprotein phenotype.

DISCUSSION

In this cross-sectional cohort of patients with PBC, sdLDL levels were independently associated with MetS, whereas no significant associations were observed with liver fibrosis stage, MASLD, or markers of cholestatic activity. Importantly, the association between MetS and sdLDL remained stable after adjustment for age, sex, adiposity, and lipid-lowering therapy. These findings suggest that sdLDL in contemporary PBC cohorts primarily reflects superimposed metabolic dysfunction rather than cholestatic liver injury itself.

Hypercholesterolemia is a well-recognized feature of PBC and affects the majority of patients during the course of disease[29]. At first glance, markedly elevated total cholesterol and LDL-C levels may suggest a classical atherogenic lipid profile. However, lipid alterations in cholestatic liver disease fundamentally differ from those observed in metabolic dyslipidemia. Impaired bile secretion promotes accumulation of phospholipids and free cholesterol in plasma and leads to formation of LpX, an abnormal phospholipid-rich lipoprotein particle largely lacking ApoB[30]. Unlike native LDL particles, LpX is not cleared through LDL receptors and demonstrates distinct structural and functional properties[31]. Mechanistically, LpX formation is linked to impaired biliary phospholipid transport and altered lipid remodeling, including reduced lecithin-cholesterol acyltransferase activity[32]. Importantly, experimental and clinical observations suggest that LpX-containing fractions may be less susceptible to oxidative modification and may attenuate LDL atherogenicity[33]. Consistent with this concept, systematic reviews and cohort studies have not demonstrated a clear increase in cardiovascular morbidity or mortality in PBC despite frequently elevated cholesterol concentrations[34]. Contemporary evidence therefore suggests that cardiovascular risk in PBC is driven predominantly by traditional metabolic risk factors rather than cholestasis-associated hypercholesterolemia alone[29]. Because the majority of patients received contemporary guideline-based therapy including UDCA, our findings likely reflect lipid remodeling patterns in treated modern PBC cohorts rather than untreated advanced cholestatic disease.

In contrast to this cholestasis-driven lipid phenotype, our findings indicate that sdLDL is linked predominantly to metabolic dysregulation. Formation of sdLDL is tightly associated with insulin resistance and hypertriglyceridemia. As described by Ginsberg and others, insulin-resistant states promote hepatic overproduction of large triglyceride-rich VLDL1 particles, which subsequently undergo cholesteryl ester transfer protein-mediated lipid exchange with LDL particles[11]. Triglyceride-enriched LDL particles then become preferred substrates for hepatic lipase, resulting in formation of smaller and denser LDL particles characteristic of the sdLDL phenotype[35]. Importantly, this remodeling cascade depends on a triglyceride-rich metabolic environment and does not arise from impaired bile excretion or cholestatic lipid retention. Therefore, the absence of association between sdLDL and cholestatic markers in our cohort is mechanistically consistent with current understanding of lipid metabolism.

Beyond their formation, sdLDL particles possess several properties that increase their atherogenic potential. Compared with large buoyant LDL particles, sdLDL demonstrates reduced affinity for LDL receptors, prolonged plasma residence time, greater susceptibility to oxidative modification, and enhanced penetration into the arterial intima[35]. These properties explain why sdLDL is increasingly regarded as a qualitative marker of atherogenic dyslipidemia rather than merely a quantitative LDL-C abnormality.

One of the most notable findings of our study was the paradoxical relationship between conventional LDL-C and metabolically driven lipoprotein remodeling. Patients with MetS exhibited significantly lower LDL-C and large LDL levels despite simultaneously demonstrating higher sdLDL concentrations and increased sdLDL/LargeLDL ratios. Similarly, LDL-C and large LDL showed inverse discriminatory performance in ROC analyses, whereas sdLDL-related parameters demonstrated modest but significant discriminatory ability for identification of MetS. However, the observed AUC values indicate limited diagnostic accuracy and should be interpreted primarily as evidence of biological association rather than clinical classification performance. Collectively, these findings support the concept that conventional LDL-C measurements may inadequately reflect qualitative atherogenic lipoprotein remodeling in metabolically affected PBC patients.

Although MetS was the most consistent independent correlate of ln(sdLDL), the explanatory power of the model was modest. Therefore, our findings should not be interpreted as indicating that MetS quantitatively determines sdLDL levels in PBC. Rather, MetS appears to identify a clinical phenotype characterized by qualitative lipoprotein remodeling toward smaller and denser LDL particles. MetS is defined as a cluster of metabolic abnormalities. It cannot be ruled out that sdLDL levels are associated with individual components of MetS. However, to avoid collinearity, we did not include individual metabolic abnormalities in our analyses, instead, we used MetS as a single variable. This interpretation is supported by the concurrent pattern of lower LDL-C and large LDL levels, higher sdLDL levels, increased sdLDL/LargeLDL ratio, and inverse discriminatory performance of LDL-C and large LDL in ROC analyses.

Interestingly, ApoB concentrations did not differ significantly according to MetS status despite substantial differences in sdLDL levels and lipoprotein remodeling indices. This observation may suggest that qualitative alterations in LDL particle composition occur even in the absence of major differences in total ApoB-containing lipoprotein burden. In this context, sdLDL-related parameters may capture aspects of metabolically driven lipoprotein remodeling not fully reflected by conventional ApoB measurements.

Our findings further reinforce the concept that lipid abnormalities in PBC are heterogeneous. While cholestasis is characterized predominantly by LpX accumulation and a distinct hypercholesterolemic phenotype, sdLDL appears to reflect a superimposed metabolic phenotype rather than an intrinsic feature of cholestatic liver disease. This mechanistic separation may partly explain why cardiovascular risk in classical PBC without metabolic comorbidity is not consistently increased, whereas patients with concomitant metabolic dysfunction may exhibit a more atherogenic lipoprotein profile. These findings do not imply that sdLDL measurement should replace standard cardiovascular risk assessment. However, advanced lipoprotein profiling may provide complementary information and help distinguish metabolically driven atherogenic dyslipidemia from cholestasis-associated hypercholesterolemia in patients with PBC (Figure 4). To our knowledge, this is the first study evaluating sdLDL as a marker of metabolic dysfunction in patients with PBC.

Figure 4 Mechanistically distinct lipid remodeling pathways in primary biliary cholangitis. Left panel: Cholestasis impairs bile acid secretion and disrupts biliary phospholipid transport, leading to reflux of phospholipids and free cholesterol into the circulation and reduced lecithin-cholesterol acyltransferase activity. These alterations promote the formation of lipoprotein X, a phospholipid-rich, apolipoprotein B (ApoB)-deficient particle that contributes to hypercholesterolemia but does not reflect classical ApoB-driven atherogenic dyslipidemia. Right panel: In contrast, insulin resistance associated with metabolic syndrome promotes hepatic overproduction of triglyceride-rich very low-density lipoprotein 1 particles. Subsequent cholesteryl ester transfer protein-mediated lipid exchange and hepatic lipase-dependent remodeling generate small dense low-density lipoprotein (sdLDL), an ApoB-containing lipoprotein subfraction characterized by prolonged circulation time, increased oxidative susceptibility, and enhanced arterial wall penetration. These pathways are mechanistically distinct and support the concept that sdLDL levels in primary biliary cholangitis primarily reflect superimposed metabolic dysfunction rather than intrinsic cholestatic activity. Created by the authors for illustrative purposes (OpenAI 1.3.2026). LCAT: Lecithin-cholesterol acyltransferase; LpX: Lipoprotein X; PBC: Primary biliary cholangitis; sdLDL: Small dense low-density lipoprotein; LDL: Low-density lipoprotein; VLDL: Very low-density lipoprotein; ApoB: Apolipoprotein B; CETP: Cholesteryl ester transfer protein; TG: Triglycerides.

Interestingly, sdLDL levels were associated with MetS but not with MASLD. Although MASLD was present in a substantial proportion of patients, sdLDL concentrations did not significantly differ according to steatosis status. This discrepancy likely reflects conceptual differences between MASLD and MetS. According to current consensus definitions, MASLD requires hepatic steatosis together with at least one cardiometabolic risk factor[36]. However, this definition encompasses a heterogeneous population ranging from mild metabolic disturbance to overt insulin resistance. In contrast, MetS represents a more advanced systemic metabolic phenotype characterized by central obesity, insulin resistance, hypertriglyceridemia, and low HDL-C levels.

Importantly, hepatic steatosis assessed using CAP primarily quantifies liver fat accumulation and may not directly capture systemic insulin resistance or the qualitative lipoprotein remodeling required for sdLDL generation. Previous studies have shown that CAP correlates with hepatic steatosis but demonstrates weaker associations with systemic metabolic dysregulation and atherogenic lipoprotein phenotype[37,38]. Our findings therefore suggest that in PBC, sdLDL reflects systemic metabolic dysfunction rather than steatosis alone.

In univariable analyses, patients who achieved CBR demonstrated higher sdLDL concentrations. Although this finding may initially appear paradoxical, one possible explanation is that improvement of cholestasis reduces circulating LpX levels, thereby unmasking metabolically driven ApoB-containing lipoprotein remodeling. In this scenario, the lipid profile may shift from a cholestasis-dominated phenotype toward a more classical metabolic dyslipidemia in which sdLDL becomes more apparent. However, this association did not persist after multivariable adjustment, suggesting that the observed difference is likely confounded by metabolic factors rather than representing an independent effect of biochemical response. Therefore, this observation should be interpreted cautiously and viewed primarily as a hypothesis-generating finding requiring further validation in longitudinal studies.

Limitations

The present study should be interpreted in light of several limitations. First, its cross-sectional design precludes causal inference, and the observed association between MetS and sdLDL therefore reflects correlation rather than temporal sequence.

Second, this was a single-center study conducted at a tertiary referral center, which may limit generalizability. Although the cohort reflects real-world hepatology practice, external validation in independent populations is warranted. Despite the single-center design, the study included a comprehensively phenotyped PBC cohort with standardized Lipoprint lipoprotein profiling and VCTE assessment. In addition, the sample size was moderate and the proportion of male patients was low, consistent with the well-established female predominance of PBC observed in European epidemiological cohorts, but limiting adequately powered sex-specific analyses.

Third, cardiovascular outcomes were not assessed. Although sdLDL is widely regarded as a marker of atherogenic dyslipidemia, we did not evaluate incident cardiovascular events or surrogate markers of atherosclerosis. Therefore, the clinical implications of elevated sdLDL concentrations in this specific population remain mechanistic rather than outcome-driven.

Fourth, sdLDL was measured at a single time point. Lipoprotein subfractions may vary according to metabolic status, therapeutic interventions, and disease activity, and longitudinal studies are needed to evaluate temporal stability and prognostic significance. In addition, sdLDL values demonstrated a partially zero-inflated distribution, which may limit conventional linear modeling approaches despite complementary logistic analyses yielding consistent results.

Fifth, direct quantification of LpX was not performed. Therefore, its contribution to observed lipid patterns remains indirect and mechanistic. In addition, coexisting autoimmune disorders commonly associated with PBC, such as Sjögren syndrome or autoimmune thyroid disease, were not systematically analyzed and their potential influence on lipid metabolism cannot be excluded.

In addition, information on concomitant autoimmune diseases was not collected in our study, therefore, their potential effect on lipid metabolism was not evaluated and cannot be excluded.

Finally, the explanatory power of the multivariable model was modest (adjusted R² = 0.054), indicating that a substantial proportion of variability in sdLDL remains unexplained. However, sdLDL metabolism is biologically complex and influenced by multiple interacting metabolic, genetic, and environmental factors. Dietary habits and physical activity were not systematically assessed and residual lifestyle-related confounding cannot be excluded.

CONCLUSION

In conclusion, sdLDL levels in patients with PBC were independently associated with MetS while associations with cholestatic activity, liver fibrosis, MASLD, and biochemical response were not independently sustained after multivariable adjustment. These findings support the concept that sdLDL reflects superimposed metabolic dysfunction rather than intrinsic cholestatic lipid alterations in contemporary PBC cohorts. sdLDL-related parameters may help characterize a subgroup of PBC patients with metabolically driven qualitative lipoprotein remodeling beyond conventional LDL-C measurements.