Longitudinal changes in cardiometabolic risk factors are associated with changes in liver stiffness in adults with MASLD

Abstract

Background & Aims

Metabolic dysfunction-associated steatotic liver disease (MASLD) is characterized by cardiometabolic risk factors (CMRFs). Treatment of CMRFs is the cornerstone of MASLD management, yet other than weight loss, the association between changes in CMRFs and changes in liver-related measures is not defined. This study aimed to characterize these longitudinal associations using non-invasive measurements of hepatic fibrosis and steatosis.

Methods

Adult participants from the NASH CRN prospective study with biopsy-proven MASLD and at least two sequential vibration-controlled transient elastography (VCTE) exams (≥1 year apart) were included. Annual changes in CMRF parameters were compared against annual changes in liver stiffness measurement (LSM) and controlled attenuation parameter (CAP) using adjusted mixed-effects multinomial logistic regression. The primary outcome was LSM change ≥25% (compared to a reference group of no change, or ≥25% LSM change), while secondary outcomes were CAP change ≥30% and changes in histologic grading of steatohepatitis or fibrosis.

Results

A total of 1,417 participants were included (mean age 52 years, 38% male, 82% Caucasian). Annual worsening in all CMRFs (except triglycerides and HDL) was associated with LSM worsening (17% greater risk per 1% increase in hemoglobin A1c; 24% greater risk per 5% increase in weight). Improvements in all CMRFs were associated with LSM improvement (32% greater chance per 1% decrease in hemoglobin A1c; 27% greater chance per 5% decrease in weight). Similar associations were observed for changes in CAP.

Conclusions

Annual changes in CMRFs were associated with clinically relevant changes in LSM, CAP, and liver histology. These findings suggest that treatment of metabolic comorbidities is associated with the trajectory of MASLD.

Graphical Abstract

Introduction

Metabolic dysfunction-associated steatotic liver disease (MASLD) encompasses a spectrum from benign steatosis to advanced fibrosis and is defined by the presence of steatosis and at least one of five cardiometabolic risk factors (CMRFs).^1,2 The co-existence of these CMRFs with MASLD is well documented and the disorders associated with individual CMRFs, such as type 2 diabetes, are associated with fibrosis progression, hepatic decompensation, and liver-related mortality.^3–6 The influence of these individual CMRFs is thus relevant to both liver-related trajectories and extrahepatic outcomes, as overall mortality in individuals with MASLD is driven by cardiovascular disease and extrahepatic malignancy, as well as progressive liver fibrosis.^3,7,8 Understanding of the impact of longitudinal changes in CMRFs on the natural history of MASLD is largely based on retrospective observational data, which show that improvements in hemoglobin A1c (HbA1c) achieved with metabolic therapies are associated with improvements in non-invasive markers of hepatic fibrosis, and that relative weight loss of 5–10% can reverse steatosis and steatohepatitis.^9,10 Use of non-invasive assessments of hepatic steatosis and fibrosis, through vibration-controlled transient elastography (VCTE, or FibroScan^®), is important for disease monitoring and as a surrogate therapeutic endpoint, as serial monitoring of fibrosis via liver biopsy is not feasible in general clinical practice.^2,11,12 Given the slow and heterogenous progression of MASLD to advanced fibrosis and cirrhosis, understanding the impact of longitudinal changes in the associated risk factors is critical for therapeutic management, risk stratification, and clinical prognostication. Furthermore, while optimization of cardiometabolic comorbidities is a key feature of MASLD treatment guidelines, there is limited evidence connecting this optimization with changes in liver measurements and little guidance as to the exact threshold of “optimization” that is associated with liver benefits, particularly among the CMRFs that are used to define the presence of MASLD. Understanding the impact of these CMRF changes is also important as pharmacotherapy used to induce these changes may lead to improvements in liver stiffness independent of their CMRF effects. This study aims to utilize a large prospective trial of participants with biopsy-proven MASLD to explore how the trajectory of changes in cardiometabolic parameters used to define MASLD are associated with changes in non-invasive measurements of hepatic fibrosis and steatosis.

Patients and methods

Study population

This study included participants 18 years of age or older with biopsy-proven MASLD enrolled in the multicenter NASH Clinical Research Network (NASH CRN) prospective observational NAFLD Database 2 (ClinicalTrials.Gov identifier: NCT01030484) and NAFLD Database 3 (ClinicalTrials.Gov identifier: NCT04454463) studies who had VCTE exams between 2014 and 2023 at nine clinical centers in the United States. The NAFLD Database 2 study was approved by the institutional review board (IRB) at each participating center and the data coordinating center, and the NAFLD Database 3 study was approved by the single IRB at Johns Hopkins School of Medicine for all participating centers and the data coordinating center. Participants provided written informed consent. Participants were included if they had at least two valid VCTE exams separated by at least 1 year. Participants were excluded from the analysis if they did not have a prior diagnosis of MASLD on biopsy or if they were missing CMRF parameters at the first VCTE exam. A CONSORT diagram for included participant flow is shown in Fig. S1.

Study visits and procedures

Study visits were performed annually. Study site investigators were instructed to provide counseling regarding lifestyle modifications relevant to MASLD management in accordance with standardized guidance from the central study steering committee. This guidance included dietary recommendations (such as low saturated fat intake, low fructose and sugar-sweetened beverage intake, and targeting a hypocaloric diet), individualized weight loss targeting 5-10% body weight loss, abstinence from alcohol, at least 150 min of moderate-intensity aerobic exercise per week, and smoking cessation. Participants completed anthropometric measurements, physical examination, study-specific questionnaires, and blood laboratory analysis at each study visit. Routine laboratory analysis was performed at each clinical site according to the study protocols. Demographics were obtained at trial enrollment. VCTE exams were performed at annual study visits following overnight fasting using FibroScan^® 502 Touch or FibroScan^® 630 Expert (Echosens, Paris, France). Valid VCTE exams were defined by at least 10 sequential valid measurements for liver stiffness measurement (LSM) with the LSM IQR/ median ≤30%. FibroScan^® examination was performed using either an M+ or XL+ probe as suggested by the device software automatic probe selection tool. All examinations were performed by a trained and certified study coordinator at each site using a standardized protocol, following supervised didactic and hands-on device training.^13,14

Covariates of interest

CMRF parameters of interest were aligned with the CMRFs utilized in the consensus definition of MASLD: BMI, body weight, waist circumference, HbA1c, systolic blood pressure, diastolic blood pressure, triglycerides, and HDL-cholesterol.¹ The primary exposure of interest was the annual change in each of these parameters, relative to the previous value. Medication use was obtained from participants at each study visit. Changes in CMRF parameters were reported as “worsening” if the net change was an increase (or decrease if HDL) or “improving” if the net change was a decrease (or increase if HDL).

LSM and controlled attenuation parameter (CAP) were obtained from each valid VCTE exam (or the first valid exam if multiple valid exams were performed at a single study visit) per participant.

Outcomes

The primary outcome of interest was change in LSM by VCTE, relative to previous sequential annual VCTE examination. This outcome was classified in a three-level categorical variable as i) worsened by ≥25% relative to previous value; ii) improved by ≥25% relative to previous value; or iii) change within ± 25% of previous value (i.e. “no change”). This threshold, along with the use of relative percentage change in LSM, was selected to align with recent clinical practice guidance on clinically meaningful LSM reductions assessed by VCTE following treatment with MASLD-specific therapeutics.^15,16

Secondary outcomes of interest included change in liver fat content measured by CAP, relative to previous sequential annual VCTE examination. This outcome was categorized as i) worsened by ≥30% relative to previous value; ii) improved by ≥30% relative to previous value; or iii) change within ≥30% of previous value (i.e. “no change”). This threshold was selected to align with recent expert guidance recommendations for a clinically meaningful hepatic steatosis reduction (assessed by MRI proton-derived fat fraction) following treatment with metabolic dysfunction-associated steatohepatitis (MASH)-specific therapeutics.¹⁶

Additional secondary outcomes of interest included changes in histologic parameters between baseline and follow-up liver biopsy (if available). There was no protocolized requirement for liver biopsy(ies) other than baseline biopsy required for study enrollment and thus any subsequent biopsies were obtained based on clinical indication at the discretion of individual treating clinicians. The primary histologic outcome assessed was improvement in NAFLD activity score (NAS) by ≥2 points with no worsening of histologic fibrosis stage.¹⁷ The second histologic outcome assessed was improvement in histologic fibrosis stage by 1 stage or greater (range 0-4) with no worsening of histologic MASH diagnosis (e.g. a diagnosis of not MASH to MASH). Histologic criteria for NAS, steatohepatitis, and fibrosis staging are shown in Tables S1 and S2.

Statistical analysis

Categorical variables were expressed as frequency and percentage of total, while continuous variables were expressed as mean (± standard deviation) or median (interquartile range).

Analyses of the primary endpoint of relative change in LSM and the secondary endpoint of relative change in CAP were performed using mixed-effects multinomial logistic regression analysis for repeated measures with the outcome categorized as improvement, no change, or worsening, with the no change category used as the baseline reference category in our analysis. The annual change of each individual CMRF parameter was assessed as a repeated measure in this model and unit changes were scaled to clinically relevant and feasible values to align with relevant treatment recommendation guidelines where possible.^18–20 Use of these scaled values, rather than single unit changes, alters the magnitude of model estimates but does not affect the strength of statistical association (p value).

Each statistical model included terms for baseline value of the CMRF parameter as well as time. These models were adjusted for age, sex, race, ethnicity, BMI at each time point, baseline value of the analyzed CMRF parameter, and relevant medications (HbA1c model adjusted for diabetes medication use, blood pressure models adjusted for antihypertensive use, and triglycerides and HDL models adjusted for statin use). Diabetes medications include glucagon-like peptide 1 receptor agonists (GLP-1s), insulin, metformin, sulfonylureas, meglitinides, thiazolidinediones, or other oral antihyperglycemic agents. Antihypertensives include calcium channel blockers, beta blockers, angiotensin converting enzyme inhibitors (ACE-inhibitors), amgiotensin II receptor blockers, or oral diuretics. The models utilizing body weight or waist circumference as the parameter of interest were not adjusted for BMI at each time point. Relative risk ratios (RRRs) are interpreted as the relative risk of LSM or CAP worsening or improvement, vs. no change in outcome, per defined unit change in CMRF parameter.

These results are reported in separate tables for each level of outcome (worsening vs. no change and improvement vs. no change) to allow reporting of CMRF parameters as clinically interpretable changes, such as reporting the increase in a CMRF parameter with the “worsening” outcome and a decrease in the CMRF parameter with the “improvement” outcome.

In a sensitivity analysis, marginal structural models with stabilized inverse probability of treatment weights were used to account for potential time-dependent confounding in CMRF-specific pharmacotherapy treatment (statins, blood pressure medications including ACE-inhibitors, and diabetes medications including GLP-1 inhibitors) and censoring.²¹ Specifically, past values of a CMRF parameter influence the decision to prescribe a specific pharmacotherapy, which may then influence future CMRF parameter values or may directly influence liver disease progression. To generate inverse probability of treatment weights for each CMRF-specific pharmacotherapy, pooled logistic regression models were used to estimate (1) the time-varying probability of each treatment (or no treatment) and (2) the time-varying probability of study discontinuation (censoring). These probability models included both time-invariant baseline factors and time-varying factors as independent variables to estimate the denominator of each weight (Table S3). For stabilized weights, the model for the numerator of each weight included time-invariant baseline factors as independent variables. Final stabilized weights were calculated for single medications as (treatment probability numerator/treatment probability denominator) * (censoring probability numerator/censoring probability denominator) and multiple drugs as (treatment 1 numerator/treatment 1 denominator) * (treatment 2 numerator/treatment 2 denominator) * (censoring numerator/censoring denominator). We reviewed the distribution of the stabilized weights and excluded observations with extreme weights (<5th or >95th percentile) from the weighted analyses. Balance in variables by treatment was assessed using pbalchk in Stata, confirming balance improved with the use of the stabilized weights. Demographic, comorbidity, and medication variables remaining imbalanced (standardized difference >0.1) were included as covariates in the multivariable marginal structural models. The final marginal structural multinomial logistic regression models were weighted using stabilized inverse probability of treatment weights and adjusted for the original model covariates, as well as any covariates that remained imbalanced after weighting.

Analyses of the secondary histologic outcomes (NAS improvement with no worsening of fibrosis or fibrosis improvement with no worsening of MASH) were performed with logistic regression analysis, using the binary outcome variable of improvement vs. no improvement. This model was used to assess each CMRF parameter individually. These models were adjusted for age, sex, race, ethnicity, BMI, baseline value of the histologic outcome, time between biopsies, baseline value of the CMRF parameter, and relevant medications (HbA1c model adjusted for diabetes medication use, blood pressure models adjusted for antihypertensive use, and triglycerides and HDL models adjusted for statin use). Odds ratios are interpreted as the odds of outcome (histologic improvement) per defined unit change in CMRF parameter.

Analyses were performed using Stata release 18.²² All statistical tests were two-tailed, and results were deemed to be statistically significant if p <0.05.

Results

Participant characteristics

A total of 1,417 participants met the inclusion criteria (Table 1, Fig. S1). This population had a mean age of 51.8 years (±12.1 years), was 38% male, 82% Caucasian, and 11% were of Hispanic ethnicity. The baseline mean body weight was 96.7 kg (±21.6 kg), mean waist circumference was 108.9 cm (±15.0 cm), and mean BMI was 34.1 kg/m² (±6.6 kg/m²), with 39% having a BMI ≥35 kg/m². Metabolic comorbidities were common, with 52% having type 2 diabetes, 60% hypertension, and 58% metabolic syndrome, while 15% were current regular cigarette smokers and 51% reported never drinking alcohol. Baseline CMRF parameters included a mean HbA1c of 6.4% (±1.3%), mean systolic blood pressure of 132.0 mmHg (±15.5 mmHg), mean diastolic blood pressure of 78.6 mmHg (±10.1 mmHg), mean HDL of 45.8 mg/dl (±12.6 mg/dl), and median triglycerides of 139.0 mg/dl (104.0–200.5 mg/dl).

Table 1.

Baseline characteristics of study population.

Participant characteristic	N = 1,417
Demographics
Age (years), mean (SD)	51.8 (12.1)
Sex
Female	879 (62%)
Male	538 (38%)
Caucasian race	1,165 (82%)
Hispanic ethnicity	157 (11%)
Physical exam
Weight (kg), mean (SD)	96.7 (21.6)
Waist circumference (cm), mean (SD)	108.9 (15.0)
BMI (kg/m2), mean (SD)	34.1 (6.6)
BMI categories
<25 kg/m2	69 (5%)
25-29 kg/m2	348 (25%)
30-34 kg/m2	450 (32%)
≥35 kg/m2	550 (39%)
Systolic blood pressure (mmHg), mean (SD)	132.0 (15.5)
Diastolic blood pressure (mmHg), mean (SD)	78.6 (10.1)
Smoking and alcohol consumption
Current “regular” smoker	219 (15%)
Alcohol consumption
Never	726 (51%)
Monthly or less	427 (30%)
2-4 times/month	157 (11%)
2-3 times/week	75 (5%)
≥4 times/week	32 (2%)
Comorbidities
Type 2 diabetes	740 (52%)
Hypertension	844 (60%)
Coronary artery disease	58 (4%)
Metabolic syndrome	818 (58%)
Chronic kidney disease	37 (3%)
Laboratory values
ALT (U/L), median (IQR)	44.0 (29.0–67.0)
AST (U/L), median (IQR)	35.0 (25.0–51.0)
Total bilirubin, mg/dl	0.7 (0.4)
Albumin, g/dl, median (IQR)	4.4 (4.2–4.6)
HbA1c (%), mean (SD)	6.4 (1.3)
HOMA-IR, mean (SD)	8.2 (11.9)
Total cholesterol (mg/dl), mean (SD)	179.4 (41.6)
HDL cholesterol (mg/dl), mean (SD)	45.8 (12.6)
LDL cholesterol (mg/dl), mean (SD)	103.0 (36.0)
Triglycerides (mg/dl), median (IQR)	139.0 (104.0–200.5)
Medication use
Any diabetes medication	645 (46%)
Metformin	503 (38%)
Insulin	166 (12%)
Sulfonylurea	149 (11%)
GLP-1	98 (7%)
SGLT2 inhibitor	44 (3%)
Thiazolidinedione	17 (1%)
Statin	566 (40%)
ACE inhibitor	348 (25%)
VCTE
Number of VCTE exams/participant, mean (SD)	4.6 (2.4) (range 2-11)
LSM, kPa
Mean (SD)	11.2 (10.6)
Median (IQR)	7.8 (5.4–12.5)
Baseline LSM, kPa
<10 kPa	906 (63.9%)
≥10 kPa to <15 kPa	251 (17.7%)
≥15 kPa	260 (18.4%)
CAP, dB/m
Mean (SD)	316.6 (55.0)
Median (IQR)	325.0 (285.0–356.0)
Liver biopsy
NAS, baseline, mean (SD)	4.3 (1.6)
Fibrosis Stage, baseline, mean (SD)	1.8 (1.3)

ACE, angiotensin converting enzyme; ALT, alanine aminotransferase; AST, aspartate aminotransferase; CAP, controlled attenuation parameter; GLP-1, glucagon-like peptide 1; HbA1c, hemoglobin A1c; LSM, liver stiffness measurement; NAS, NAFLD activity score; SGLT2, sodium-glucose transport protein 2; VCTE, vibration-controlled transient elastography.

A mean of 4.6 (±2.4, range 2–11) annual VCTE exams were performed per participant, with a median LSM of 7.8 kPa (5.4–12.5 kPa) and a median CAP of 325.0 dB/m (285.0–356.0 dB/m). The median time between sequential VCTE exams was 1.0 years (0.9–1.3 years). The median total duration of disease captured by VCTE (median time between first and last VCTE exam) was 4.3 years (2.3–7.1 years). There were 95 participants with paired liver biopsies with a mean time of 2.9 years (±1.8 years) between biopsies.

Association of changes in CMRF parameters with longitudinal changes in LSM

A total of 889 participants (62.7%) had at least one VCTE exam with LSM worsening by ≥25% relative to the previous exam, 885 participants (62.5%) had LSM improvement by ≥25%, and 626 participants (44.2%) experienced both a period of improvement and a period of worsening between paired exams (participants could have both if they had at least three VCTE exams). A total of 1,089 participants (77%) had at least one VCTE exam with LSM change within ±25% of the previous value (i.e. “no change”). Worsening of LSM was significantly associated with worsening of each analyzed CMRF parameter, except triglycerides and HDL (Table 2, Fig. 1). In adjusted longitudinal mixed-effects multinomial logistic regression analysis, the risk of LSM worsening (≥25% increase relative to previous exam, vs. no change in LSM) was associated with an increase in HbA1c (per 1% increase, adjusted RRR [aRRR] 1.17, 95% CI 1.09-1.27, p <0.001), an increase in BMI (per 1 kg/m² increase, aRRR 1.08, 95% CI 1.04-1.12, p <0.001), a relative increase in body weight (per 5% increase, aRRR 1.24, 95% CI 1.16-1.33, p <0.001), an increase in waist circumference (per 5 cm increase, aRRR 1.10, 95% CI 1.05-1.16, p <0.001), an increase in systolic blood pressure (per 10 mmHg increase, aRRR 1.10, 95% CI 1.06-1.16, p <0.001), and an increase in diastolic blood pressure (per 10 mmHg increase, aRRR 1.11, 95% CI 1.04-1.19, p = 0.003). The risk of LSM worsening (vs. no change in LSM) was nearly significantly associated with a relative increase in triglycerides (per 20% increase, aRRR 1.04, 95% CI 1.00-1.08, p = 0.05), and was not associated with a decrease in HDL (per 10 mg/dl decrease, aRRR 1.06, 95% CI 0.97-1.16, p = 0.18).

Table 2.

Association between worsening in CMRF parameters and ≥25% worsening in LSM.

	LSM worsened by 25% vs. no change
	aRRR (95% CI)†	p values
Change in HbA1c (per 1% increase)	1.17 (1.09–1.27)	<0.001
Change in BMI (per 1 kg/m2 increase)	1.08 (1.04–1.12)	<0.001
Relative change in body weight (kg) (per 5% increase)	1.24 (1.16–1.33)	<0.001
Change in waist circumference (per 5 cm increase)	1.10 (1.05–1.16)	<0.001
Change in systolic blood pressure (per 10 mmHg increase)	1.10 (1.06–1.16)	<0.001
Change in diastolic blood pressure (per 10 mmHg increase)	1.11 (1.04–1.19)	0.003
Relative change in triglycerides (mg/dl) (per 20% increase)	1.04 (1.00–1.08)	0.05
Change in HDL (per 10 mg/dl decrease)	1.06 (0.97–1.16)	0.18

aRRR, adjusted relative risk ratio; CMRF, cardiometabolic risk factor; HbA1c, hemoglobin A1c; LSM, liver stiffness measurement.

Bold denotes statistically significant changes at α = 0.05.

^†From adjusted longitudinal mixed-effects multinomial logistic regression analysis, adjusted for age, sex, race, ethnicity, BMI at each time point, and baseline value of the analyzed CMRF. Adjustment for BMI at each time point was not performed in the models for BMI, body weight, or waist circumference. HbA1c model adjusted for diabetes medications, systolic and diastolic blood pressure models were adjusted for use of antihypertensive medications, and triglycerides and HDL models were adjusted for statin use.

Fig. 1. Association between worsening in CMRF parameters and ≥25% worsening in LSM.

This figure shows the aRRR for the association between changes in CMRF parameters and risk of LSM worsening by ≥25% vs. no change in LSM. Bold text denotes statistically significant p values at α = 0.05. aRRR, adjusted relative risk ratio; BP, blood pressure; CMRF, cardiometabolic risk factor; HbA1c, hemoglobin A1c; LSM, liver stiffness measurement.

Improvement in LSM was significantly associated with improvements in each analyzed CMRF parameter (Table 3, Fig. 2). In adjusted longitudinal mixed-effects multinomial logistic regression analysis, the chance of LSM improvement (≥25% decrease relative to previous exam, vs. no change in LSM) was associated with a decrease in HbA1c (per 1% decrease, aRRR 1.32, 95% CI 1.20-1.43, p <0.001), a decrease in BMI (per 1 kg/m² decrease, aRRR 1.12, 95% CI 1.08-1.16, p <0.001), a relative decrease in body weight (per 5% decrease, aRRR 1.27, 95% CI 1.19-1.37, p <0.001), a decrease in waist circumference (per 5 cm decrease, aRRR 1.09, 95% CI 1.04-1.15, p = 0.001), a decrease in systolic blood pressure (per 10 mmHg decrease, aRRR 1.05, 95% CI 1.01-1.10, p = 0.01), a decrease in diastolic blood pressure (per 10 mmHg decrease, aRRR 1.09, 95% CI 1.01-1.16, p = 0.03), a relative decrease in triglycerides (per 20% decrease, aRRR 1.09, 95% CI 1.04-1.12, p <0.001), and an increase in HDL (per 10 mg/dl increase, aRRR 1.20, 95% CI 1.09-1.31, p <0.001).

Table 3.

Association between improvement in CMRF parameters and ≥25% improvement in LSM.

	LSM improved by 25% vs. no change
	aRRR (95% CI)†	p values
Change in HbA1c (per 1% decrease)	1.32 (1.20–1.43)	<0.001
Change in BMI (per 1 kg/m2 decrease)	1.12 (1.08–1.16)	<0.001
Relative change in body weight (kg) (per 5% decrease)	1.27 (1.19–1.37)	<0.001
Change in waist circumference (per 5 cm decrease)	1.09 (1.04–1.15)	0.001
Change in systolic blood pressure (per 10 mmHg decrease)	1.05 (1.01–1.10)	0.01
Change in diastolic blood pressure (per 10 mmHg decrease)	1.09 (1.01–1.16)	0.03
Relative change in triglycerides (mg/dl) (per 20% decrease)	1.09 (1.04–1.12)	<0.001
Change in HDL (per 10 mg/dl increase)	1.20 (1.09–1.31)	<0.001

aRRR, adjusted relative risk ratio; CMRF, cardiometabolic risk factor; HbA1c, hemoglobin A1c; LSM, liver stiffness measurement.

Bold denotes statistically significant changes at α = 0.05.

^†From adjusted longitudinal mixed-effects multinomial logistic regression analysis, adjusted for age, sex, race, ethnicity, BMI at each time point, and baseline value of the analyzed CMRF. Adjustment for BMI at each time point was not performed in the models for BMI, body weight, or waist circumference. HbA1c model adjusted for diabetes medications, systolic and diastolic blood pressure models were adjusted for use of antihypertensive medications, and triglycerides and HDL models were adjusted for statin use.

Fig. 2. Association between improvement in CMRF parameters and ≥25% improvement in LSM.

This figure shows the aRRR for the association between changes in CMRF parameters and likelihood of LSM improvement by ≥25% vs. no change in LSM. Bold text denotes statistically significant p values at α = 0.05. aRRR, adjusted relative risk ratio; BP, blood pressure; CMRF, cardiometabolic risk factor; HbA1c, hemoglobin A1c; LSM, liver stiffness measurement.

In a sensitivity analysis using marginal structural modeling to account for time-dependent confounding of CMRF-specific pharmacotherapy, results were largely similar to the main analysis for each CMRF parameter model (Tables S4 and S5). For the outcome of worsening LSM, the risk estimates for change in HbA1c (per 1% change, aRRR 1.16, 95% CI 1.00-1.35, p = 0.06) and change in diastolic blood pressure (per 10 mmHg, aRRR 1.12, 95% CI 0.99-1.19, p = 0.07) were the same as the main analysis but were no longer statistically significant. Change in systolic blood pressure (per 10 mmHg change, aRRR 1.20, 95% CI 1.09-1.33, p <0.001) demonstrated a greater effect size than the main analysis (aRRR 1.10, p <0.001), as did change in triglycerides (per 20% relative change, aRRR 1.10, 95% CI 1.03-1.17, p = 0.007), which became statistically significant (main model: aRRR 1.04, p = 0.05).

For the outcome of improvement in LSM, change in systolic blood pressure (per 10 mmHg decrease, aRRR 1.08, 95% CI 0.99-1.18, p = 0.07) had a similar magnitude of effect as observed in the main model (aRRR 1.05, p = 0.01) as did change in diastolic blood pressure (per 10 mmHg decrease, aRRR 1.11, 95% CI 0.96-1.27, p = 0.16) (main model aRRR 1.09, p = 0.03), but both were no longer statistically significant.

Association of changes in CMRF parameters with longitudinal changes in CAP

A total of 239 participants (16.9%) had CAP worsening by ≥30% relative to previous VCTE exam, a total of 295 participants (20.8%) had CAP improvement by ≥30% relative to previous VCTE exam, and a total of 132 participants (9.3%) had both an episode of improvement and an episode of worsening between paired exams (a participant may have had both improvement and worsening if they had at least three VCTE exams). There were a total of 1,341 participants (95%) who had at least one VCTE exam with CAP change within ±30% of the previous value (i.e. “no change”). Worsening of CAP was significantly associated with worsening of all CMRF parameters except for HDL (Table S6, Fig. 3). In adjusted longitudinal mixed-effects multinomial logistic regression analysis, the risk of CAP worsening (≥30% increase relative to previous exam, vs. no change in CAP) was significantly associated with an increase in HbA1c (per 1% increase, aRRR 1.20, 95% CI 1.05-1.37, p = 0.008), an increase in BMI (per 1 kg/m² increase, aRRR 1.22, 95% CI 1.15-1.29, p <0.001), a relative increase in body weight (kg) (per 5% increase, aRRR 1.49, 95% CI 1.28-1.72, p <0.001), an increase in waist circumference (per 5 cm increase, aRRR 1.18, 95% CI 1.08-1.29, p <0.001), an increase in systolic blood pressure (per 10 mmHg increase, aRRR 1.11, 95% CI 1.03-1.19, p = 0.004), an increase in diastolic blood pressure (per 10 mmHg increase, aRRR 1.20, 95% CI 1.06-1.36, p = 0.005), and a relative increase in triglycerides (mg/dl) (per 20% increase, aRRR 1.13, 95% CI 1.05-1.22, p = 0.002). Risk of CAP worsening (vs. no change in CAP) was not associated with a decrease in HDL (mg/dl) (per 10 mg/dl decrease, aRRR 1.05, 95% CI 0.89-1.23, p = 0.57).

Fig. 3. Association between worsening in CMRF parameters and ≥30% worsening in CAP.

This figure shows the aRRRs for the association between changes in CMRF parameters and risk of CAP worsening by ≥30% vs. no change in CAP. Bold text denotes statistically significant p values at α = 0.05. aRRR, adjusted relative risk ratio; BP, blood pressure; CAP, controlled attenuation parameter; CMRF, cardiometabolic risk factor; HbA1c, hemoglobin A1c.

Improvement of CAP by ≥30% was significantly associated with improvement of all CMRF parameters (Table 4, Fig. 4). In adjusted longitudinal mixed-effects multinomial logistic regression analysis, the chance of CAP improvement (≥30% decrease relative to previous exam, vs. no change in CAP) was associated with a decrease in HbA1c (per 1% decrease, aRRR 1.45, 95% CI 1.27-1.67, p <0.001), a decrease in BMI (per 1 kg/m² decrease, aRRR 1.28, 95% CI 1.22-1.35, p <0.001), a relative decrease in body weight (kg) (per 5% decrease, aRRR 1.67, 95% CI 1.52-1.82, p <0.001), a decrease in waist circumference (per 5 cm decrease, aRRR 1.33, 95% CI 1.20-1.47, p <0.001), a decrease in systolic blood pressure (per 10 mmHg decrease, aRRR 1.12, 95% CI 1.04-1.20, p = 0.003), a decrease in diastolic blood pressure (per 10 mmHg decrease, aRRR 1.25, 95% CI 1.12-1.41, p <0.001), a relative decrease in triglycerides (mg/dl) (per 20% decrease, aRRR 1.27, 95% CI 1.19-1.35, p <0.001), and an increase in HDL (per 10 mg/dl increase, aRRR 1.32, 95% CI 1.13-1.53, p <0.001).

Table 4.

Association between improvement in CMRF parameters and ≥30% improvement in CAP.

	CAP improved by 30% vs. no change
	aRRR (95% CI) †	p values
Change in HbA1c (per 1% decrease)	1.45 (1.27–1.67)	<0.001
Change in BMI (per 1 kg/m2 decrease)	1.28 (1.22–1.35)	<0.001
Relative change in body weight (kg) (per 5% decrease)	1.67 (1.52–1.82)	<0.001
Change in waist circumference (per 5 cm decrease)	1.33 (1.20–1.47)	<0.001
Change in systolic blood pressure (per 10 mmHg decrease)	1.12 (1.04–1.20)	0.003
Change in diastolic blood pressure (per 10 mmHg decrease)	1.25 (1.12–1.41)	<0.001
Relative change in triglycerides (mg/dl) (per 20% decrease)	1.27 (1.19–1.35)	<0.001
Change in HDL (per 10 mg/dl increase)	1.32 (1.13–1.53)	<0.001

aRRR, adjusted relative risk ratio; CAP, controlled attenuation parameter; CMRF, cardiometabolic risk factor; HbA1c, hemoglobin A1c.

Bold denotes statistically significant changes at α = 0.05.

^†From adjusted longitudinal mixed-effects multinomial logistic regression analysis, adjusted for age, sex, race, ethnicity, BMI at each time point, and baseline value of the analyzed CMRF. Adjustment for BMI at each time point was not performed in the models for BMI, body weight, or waist circumference. HbA1c model adjusted for diabetes medications, systolic and diastolic blood pressure models were adjusted for use of antihypertensive medications, and triglycerides and HDL models were adjusted for statin use.

Fig. 4. Association between improvement in CMRF parameters and ≥30% improvement in CAP.

This figure shows the aRRR for the association between changes in CMRF parameters and likelihood of CAP improvement by ≥30% vs. no change in CAP. Bold text denotes statistically significant p values at α = 0.05. aRRR, adjusted relative risk ratio; BP, blood pressure; CAP, controlled attenuation parameter; CMRF, cardiometabolic risk factor; HbA1c, hemoglobin A1c.

Association of changes in CMRF parameters with longitudinal changes in liver histology

There were 95 participants (6.7%) with paired liver biopsies with a mean time of 2.9 years (±1.8 years) between biopsies. Of these 95 participants, 25 (26%) had improvement in NAS by ≥2 points from baseline to follow-up biopsy with no worsening of fibrosis stage. In adjusted logistic regression analysis, the likelihood of NAS improvement with no worsening of fibrosis (vs. no improvement) (Table S7) was associated with a decrease in HbA1c (per 1% decrease, aOR 2.00, 95% CI 1.03-4.00, p = 0.04), a decrease in BMI (per 1 kg/m² decrease, aOR 1.43, 95% CI 1.09-1.85, p = 0.009), a relative decrease in body weight (kg) (per 5% decrease, aOR 1.92, 95% CI 1.22-3.03, p = 0.005), and a decrease in diastolic blood pressure (per 10 mmHg decrease, aOR 2.22, 95% CI 1.18-4.17, p = 0.01). The likelihood of NAS improvement was not associated with a decrease in waist circumference, a decrease in systolic blood pressure, a relative decrease in triglycerides, or an increase in HDL.

Of those with paired liver biopsies, 23 (24%) had improvement in fibrosis by ≥1 fibrosis stage without worsening of histologic MASH diagnosis. In adjusted logistic regression analysis (Table S8), the likelihood of fibrosis improvement with no worsening of MASH (vs. no fibrosis improvement) was not significantly associated with improvements in any of the analyzed CMRF parameters.

Discussion

In this prospective study of adult participants with biopsy-proven MASLD, annual changes in CMRF parameters were associated with clinically relevant changes in LSM, CAP, and liver histology. These effects are variable across the spectrum of CMRF parameters, with the greatest effects seen with the CMRF parameters reflecting glucose intolerance (hemoglobin A1c), adiposity (body weight, BMI, and waist circumference), and blood pressure, while weaker effects were seen with lipid parameters (triglycerides and HDL). These findings suggest that treatment and optimization of metabolic comorbidities is significantly associated with the trajectory of liver-related disease progression in MASLD since the observed improvements in CMRFs were likely to be related to lifestyle modifications and/or pharmacotherapy. Crucially, these associations were evident with adjustment for time-dependent confounding introduced by pharmacotherapy treatment, supporting a direct relationship between changes in CMRF parameters and LSM beyond the CMRF-independent improvements linked with specific pharmacotherapy (such as statins).

In the analysis of longitudinal LSM change, improvements in all analyzed CMRF parameters were associated with a ≥25% relative improvement in LSM. This effect was greatest with hemoglobin A1c, where a 1% improvement was associated with a 32% greater chance of LSM improvement, and body weight, where a 5% relative improvement was associated with a 27% greater chance of LSM improvement. Similar associations were seen with worsening of CMRF parameters, which were significantly associated with a ≥25% relative worsening in LSM, except the lipid parameters of triglycerides and HDL. The greatest effects were again seen with hemoglobin A1c, where a 1% annual increase was associated with a 17% greater risk of LSM worsening, and body weight, where a 5% relative worsening was associated with a 24% greater risk of LSM worsening. These associations emerge despite adjustment for longitudinal changes in body weight, which suggests that changes in these parameters are independently associated with meaningful changes in LSM that are not explained solely by weight loss. This is an important finding, as current literature to support the benefit of optimization of metabolic comorbidities on longitudinal liver health parameters is limited. LSM is validated as a surrogate for invasive measures of liver health, disease staging, and disease activity.^12,23 Longitudinal changes in VCTE-derived LSM are associated with differential risks for incident liver-related events (with regression of LSM leading to lower risk and progression of LSM leading to greater risk), independent of baseline LSM category.¹² Our results utilizing relative longitudinal changes in LSM, rather than LSM categories, allow these findings to be applied across the spectrum of MASLD, while still carrying connection to clinically meaningful liver-related outcomes. Thus, LSM is a crucial piece in the clinical management and disease monitoring of those with MASLD, and these findings have tangible clinical relevance.

We include a sensitivity analysis using a causal inference framework, a marginal structural model, to account for time-dependent confounding introduced by the use of and adjustment for pharmacotherapy in the treatment of these CMRF comorbidities. These CMRF parameters may be impacted by this confounding, where a previous parameter value influences the decision to prescribe pharmacotherapy to target that parameter, with subsequent parameter values thus influenced by that treatment and the decision to treat. Furthermore, many of these pharmacotherapy agents have purported liver-specific benefits that may be independent of their effect on CMRF treatment. We developed this model using a causal inference framework based on a proposed directed acyclic graph (Fig. S2) mapping the connections between CMRF parameters, pharmacotherapy, and liver stiffness.

Crucially, results from this model were overall similar to results from our initial analysis. Notable differences include a loss of statistical significance in the associations for (1) increase in HbA1c and LSM worsening, (2) increase in diastolic blood pressure and LSM worsening, (3) decrease in systolic blood pressure and LSM improvement, and (4) decrease in diastolic blood pressure and LSM improvement. It is likely that these losses of statistical significance are related to loss of statistical power by excluding observations with extreme inverse probability weights, given that the magnitude of point estimates remain similar and within predicted 95% confidence intervals from our initial analysis.

As hepatic steatosis is typically the initial feature of MASLD, the dynamic trajectory of hepatic steatosis may reflect changes in the severity of metabolic dysfunction in response to therapy which will impact the prognosis of MASLD. Our results found an association between improvements in all CMRF parameters and longitudinal improvements in CAP. Similar to the findings for LSM, these effects were greatest with changes in HbA1c (32% greater chance of improvement per 1% decrease) and body weight (27% greater chance of improvement per 5% relative decrease). In contrast to the LSM findings, annual changes in HDL were strongly associated with improvements in CAP as well, with a 20% greater chance for improvement in CAP per 10 mg/dl increase in HDL. Similar associations were also seen with worsening of CMRF parameters and worsening of CAP, with only decreases in HDL not associated with this outcome. Relative body weight increase (49% increased risk per 5% relative increase) and HbA1c (20% increased risk per 1% increase) were again strong associations, but the remaining parameters of adiposity also exhibited large effects, as did diastolic blood pressure (20% increased risk per 10 mmHg increase). While CAP is not a surrogate marker for disease activity and is subject to variability between exams, the persistent strength of association of these CMRFs highlights that metabolic optimization is associated with significant effects on liver health.²⁴ The association between CAP improvement and HDL increase is an important one – steatosis is known to decrease as disease progresses (as in “burnt-out” MASH), but progressive liver disease also leads to impaired lipoprotein synthesis and decreases in HDL.²⁵ This suggests that the association between improvement in CAP and increase in HDL is not due to liver disease progression and loss of steatosis, but instead reflects that optimization of these metabolic factors is truly associated with improvements in steatosis.

The findings that changes in lipid-related parameters were not as strongly associated with changes in CAP or LSM as the remaining CMRFs aligns with the current understanding of pathophysiologic mechanisms of MASLD and MASH. Obesity and insulin resistance play key roles in the development of hepatic steatosis and have been associated with progression of liver disease and with greater risk of liver-related events. Obesity leads to peripheral lipotoxicity and excess adipose tissue availability, with these risks seemingly greater with visceral distribution of adiposity. Insulin resistance is believed to be the central underlying pathophysiology leading to hepatic steatosis via lipid accumulation in hepatocytes, where peripheral insulin resistance promotes release of free fatty acids (FFAs) from adipose tissue (with greater volume of adipose tissue in obesity), as well as decreasing its utilization in skeletal muscle. The liver is responsible for the uptake of this increased FFA pool, where hepatic insulin resistance also drives energy storage via increased gluconeogenesis, lipogenesis, and altered pancreatic insulin secretion via hepatokines. Circulating lipoproteins, such as triglycerides, are generated by the liver in efforts to safely store this excess FFA pool.²⁶ The importance of insulin resistance/type 2 diabetes in the progression of liver disease is supported in epidemiologic studies, where a meta-analysis found that the presence of type 2 diabetes doubled the risk of liver-related events.²⁷ This same meta-analysis also found that elevated BMI ≥-30 kg/m² carried the second greatest risk among CMRFs for liver-related events (with a hazard ratio of 1.2).²⁷ While this meta-analysis found that many studies had limited reporting of the remaining CMRFs, other studies find that hypertension also exerts a significant risk for progression to cirrhosis in those with MASLD or a significant risk for development of cirrhosis, decompensated liver disease, or hepatocellular carcinoma when added to those with underlying type 2 diabetes.^27–29 Thus, the pathogenesis of MASLD and its progression to cirrhosis appear to be mechanistically driven by obesity and insulin resistance, while lipid-related metabolic derangements may simply be downstream markers of hepatic upregulation of lipid storage mechanisms. Our findings align with this pathophysiologic interpretation and further support the concept that not all CMRF components of MASLD carry equal risk for outcomes.

Despite the clinical relevance and practicality of non-invasive liver assessment, histology remains the gold standard for assessment of disease activity in MASLD and has historically served as the desired primary endpoint for registration trials of MASLD- and MASH-targeted therapeutics.³⁰ Our study includes a limited proportion of participants with a second follow-up liver biopsy available for paired comparison with the baseline liver biopsy, which was required for study enrollment. In the analysis of NAS improvement, which represents the histologic steatohepatitis activity associated with risk of fibrosis progression, improvement in NAS was significantly associated with improvements in most of the analyzed CMRF parameters. These findings mirrored the results using non-invasive measurement outcomes, with improvement in HbA1c, relative body weight, and diastolic blood pressure showing the greatest association with odds of NAS improvement. Similar to our other analyses, changes in triglycerides and changes in HDL were not associated with the likelihood of improvement in NAS, while systolic blood pressure change was also not associated with this outcome. Analysis of histologic fibrosis improvement found that none of the CMRF parameters were associated with the likelihood of fibrosis stage improvement. These histology results require cautious interpretation, as the decision to obtain a follow-up liver biopsy was left to the individual treating clinician and was not required by study protocol. This suggests that those participants who received a repeat biopsy may represent a different disease phenotype than those without a repeat biopsy and the ability to apply these histology outcomes to external populations is unknown. Additionally, the total duration of disease captured by paired biopsies was less (mean of 2.9 years) than the total duration of disease captured by VCTE exams (median of 4.3 years). The difference in results between the non-invasive outcomes (LSM) and histology outcomes (fibrosis improvement) may reflect this difference in duration of disease captured or may represent that VCTE may capture smaller changes in disease regression that do not lead to overt changes in fibrosis scoring. Additionally, while the timeframe of fibrosis progression from low grade to advanced fibrosis is generally greater than 10 years, there is less understanding of the rate of fibrosis regression in MASLD.³¹ Overall, the results of these histologic outcomes provide some support for the connection between LSM improvements and improvement in histologic steatohepatitis, although the small, non-representative sample size with repeat biopsy analysis limits broad interpretation of these findings.

The use of serial VCTE examinations for dynamic monitoring of liver disease progression is another unique feature of this study. The utility of serial VCTE monitoring and its concordance with histology-based assessment has been reported in a limited number of previous studies. A meta-analysis of studies in HCV-infected patients with fibrosis assessments either by VCTE or liver biopsy found similar rates of annual fibrosis progression (and mean time to cirrhosis) between studies utilizing serial VCTE examinations and those utilizing repeat liver biopsy.³² In patients with biopsy-proven MASLD, serial LSM monitoring via VCTE performs as well as paired biopsy for prediction of future liver-related events.³³ The use of repeat VCTE for LSM 10 years after baseline examination in patients with MASLD correlates well with histologic fibrosis assessment from paired biopsies performed at a similar interval.³⁴ Thus, our study adds to this growing literature by showing that the use of annual, repeated VCTE examinations reflects the dynamic changes in CMRFs that are the focus of lifestyle or therapeutic interventions.

This work provides clinically relevant results that can be feasibly incorporated into MASLD treatment guidance across the continuum of care for persons with MASLD. While current clinical practice guidelines acknowledge the importance of treatment of cardiometabolic comorbidities, particularly in a multidisciplinary approach, there is limited guidance connecting this treatment with liver-related measures or guidance regarding specific treatment targets for these comorbidities. Our results complement and enhance these guidelines by connecting the changes in these cardiometabolic comorbidities with clinically meaningful changes in liver stiffness as well as providing tangible, achievable targets for these treatments that are aligned with other specific guidance for treatment of these comorbidities.^2,18–20,35 While these results do not dramatically change the guidance to treat and optimize these comorbidities, defining their connection with practical, numerical targets may serve to encourage aggressive treatment of these comorbidities by clinicians in primary care and metabolic health settings, motivate patients to adopt this aggressive treatment, and provide a framework for future policy and guidance development. Additionally, this work does not explicitly address longitudinal changes in other CMRFs that may be associated with the metabolic dysfunction underlying MASLD, such as chronic kidney disease. Further work is needed to help define where these additional comorbidities fit into the definition of MASLD and whether the diagnostic criteria and prognostic prediction can be refined with additional or alternative metabolic parameters.

This study has limitations. Biopsy diagnosis of MASLD was required for inclusion in this study, which necessitates clinical suspicion and referral for an invasive diagnosis with subsequent study enrollment requiring referral to a specialty, tertiary-care center. These requirements mean that the included population may not be fully representative of the full spectrum of MASLD. Further validation of these results may be required in primary care or community-based settings, although it is important to note that included participants were followed observationally and any additional care for comorbidities may have been provided in non-tertiary care settings. Second, while standardized guidance for lifestyle modification counseling was provided to study investigators, detailed data on the exact modifications made by individual participants (such as dietary data, nutritionist follow-up, or exercise level) was not captured. Third, changes in cardiometabolic parameters are observational in this study. While we have utilized supplemental modeling to adjust for the effect of time-dependent confounding introduced by pharmacotherapy, this model does require specification of variables for inclusion in the probability weight for pharmacotherapy treatment or study censoring and it is possible that these weights are inadequately specified to account for all factors that influence treatment initiation and treatment response in a real-life population.

In conclusion, this prospective observational study of participants with biopsy-proven MASLD demonstrates significant association of changes in CMRFs with longitudinal, clinically relevant changes in LSM, CAP, and histologic evidence of MASH activity. While the CMRF parameters associated with glucose intolerance, adiposity, and blood pressure consistently show these effects across the assessed outcomes, the parameters of triglycerides and HDL exerted weaker or insignificant effects. These results provide a framework for clinical management of patients with MASLD and provide novel evidence to support guideline recommendations to prioritize optimizing the management of metabolic comorbidities in patients with MASLD. Additionally, these results support the concept that the CMRF components of MASLD have differential impacts, further characterize the heterogeneity of this disease process, and provide additional hypotheses for future work to identify treatment strategies that address the systemic metabolic dysfunction seen in patients with MASLD.

Highlights.

• Longitudinal changes in cardiometabolic risk factors are associated with meaningful improvements and worsening in liver stiffness.

• Changes in body weight or waist circumference and hemoglobin A1c carry the greatest effects on changes in liver stiffness.

• Similar improvements and deteriorations are observed in hepatic steatosis, with effect sizes greater than those for liver stiffness.

Impact and implications.

Guideline-directed treatment of metabolic dysfunction-associated steatotic liver disease (MASLD) is focused on optimal treatment of cardiometabolic comorbidities, but there is limited evidence on the association between outcomes of this treatment and changes in markers of liver health. The results of this study demonstrate that longitudinal changes in these cardiometabolic parameters, particularly those associated with body weight, blood pressure, and glucose intolerance, are associated with both increases and decreases in liver stiffness measurements that represent clinically meaningful changes in liver health. These results are broadly relevant to the care of individuals with MASLD, as these results can provide meaningful treatment targets for guideline development and treating clinicians, as well as provide feasible goals for impacted individuals with MASLD.

Abbreviations

CAP

controlled attenuation parameter

CMRF

cardiometabolic risk factor

HbA1c

hemoglobin A1c

LSM

liver stiffness measurement

MASH

metabolic dysfunction-associated steatohepatitis

MASLD

metabolic dysfunction-associated steatotic liver disease

NAFLD

nonalcoholic fatty liver disease

NAS

NAFLD activity score

OR

odds ratio

(a)RRR

(adjusted) relative risk ratio

VCTE

vibration-controlled transient elastography

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.jhep.2026.02.002.

Data availability

Study data and analytic methods may be made available to other researchers, upon reasonable request to the corresponding author.