Excess Liver Fat in Type 2 Diabetes Stratified by Intrapancreatic Fat Deposition: A Meta‐Analysis of Dual‐Organ MRI Studies

ABSTRACT

Both liver fat content (LFC) and intrapancreatic fat deposition (IPFD) are frequently increased in Type 2 diabetes mellitus (T2DM). We meta‐analysed dual‐organ MRI studies comparing LFC between individuals with T2DM and non‐diabetic controls and explored whether study‐level differences in LFC varied according to IPFD burden. Seventeen studies (2142 participants) were included. Individuals with T2DM had significantly higher LFC than non‐diabetic controls (standardised mean difference [SMD] 0.89, 95% CI 0.60–1.18). In exploratory subgroup analyses, liver fat excess was larger in studies with mean IPFD ≥ 10% than in those with IPFD < 10% (SMD 1.20 vs. 0.62; p = 0.04), though this finding was not consistently retained in leave‐one‐out analyses. Alternative IPFD cutoffs yielded directionally similar but non‐significant subgroup differences. Study‐level differences in IPFD were associated with differences in LFC (β = 0.383, p = 0.006) in univariable meta‐regression, but not after BMI adjustment. These findings suggest that higher IPFD is associated with greater liver fat excess in T2DM at the study level. Individual‐level analyses are now warranted to clarify this relationship.

Key Points

• Liver fat content was significantly higher in individuals with type 2 diabetes than in non‐diabetic controls.

• Greater excess liver fat was observed in studies reporting higher intrapancreatic fat deposition (≥ 10%).

• Study‐level differences in intrapancreatic and liver fat were positively associated, but this association did not persist after adjustment for BMI.

• Individual‐level studies are needed to clarify the relationship between intrapancreatic and liver fat in type 2 diabetes.

Abbreviations

BMI

body mass index

CI

confidence interval

IPFD

intrapancreatic fat deposition

LFC

liver fat content

MOOSE

Meta‐analysis Of Observational Studies in Epidemiology

MRI

magnetic resonance imaging

PRISMA

Preferred Reporting Items for Systematic Reviews and Meta‐Analyses

ROI

region of interest

SMD

standardised mean difference

T2DM

Type 2 diabetes mellitus

1. Introduction

Liver fat content (LFC) and intrapancreatic fat deposition (IPFD) are both often increased in individuals with Type 2 diabetes mellitus (T2DM) [1, 2, 3, 4]. However, imaging studies addressing this relationship have yielded inconsistent findings. While several studies have reported significant associations between IPFD and LFC [5, 6, 7], others have found no associations [8, 9]. At least in part, this discrepancy in findings may be attributed to the use of imaging modalities less accurate than magnetic resonance imaging (MRI) and to the inclusion of populations not specifically focused on individuals with T2DM. The aim of the present study was to pool data from MRI studies assessing both LFC and IPFD in individuals with T2DM versus non‐diabetic controls, quantify liver fat excess in T2DM, and examine the relationship between study‐level differences in LFC and IPFD.

2. Methods

2.1. Data Sources and Search Strategy

We designed this study according to the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) statement and Meta‐analysis Of Observational Studies in Epidemiology (MOOSE) reporting guidelines [10, 11] (Table 1). This study was based on a dataset from a previously performed systematic review (CRD420251208380) [12]. We searched PubMed and Embase databases from inception to November 15, 2025. The full search strategies for each database are detailed in the Supporting Information. We also screened the reference lists of relevant reviews and primary articles, as well as the authors' personal libraries, to ensure comprehensive coverage.

TABLE 1.

Liver fat content differences between individuals with Type 2 diabetes and non‐diabetic controls stratified by IPFD level.

	Studies, n	Patients, n	SMD	95% confidence interval	I 2, %	τ 2	Subgroup difference
Exploratory subgroup analysis
IPFD < 10%	10	1089	0.62	0.41 to 0.83	45.6	0.04	0.04
IPFD ≥ 10%	7	1053	1.20	0.53 to 1.87	92.3	0.47
Ancillary analyses
IPFD < 6%	3	275	0.72	0.04 to 1.41	33.5	0.02	0.37
IPFD ≥ 6%	14	1867	0.93	0.57 to 1.28	89.1	0.33
IPFD < 8%	8	863	0.63	0.44 to 0.82	26.0	0.02	0.06
IPFD ≥ 8%	9	1279	1.09	0.55 to 1.64	92.0	0.46
IPFD < 12%	13	1495	0.81	0.54 to 1.08	82.7	0.17	0.46
IPFD ≥ 12%	4	647	1.15	−0.28 to 2.58	94.3	0.76

Abbreviations: IPFD, intrapancreatic fat deposition; SMD, standardised mean difference.

Inclusion in this synthesis was restricted to studies meeting the following criteria: (1) observational studies including individuals with T2DM (as defined by each study) and non‐diabetic controls; (2) MRI assessment of both LFC and IPFD performed within the same study population; (3) availability of sufficient LFC and IPFD data. Only original research articles were considered eligible. If multiple publications were based on the same study population, the most comprehensive article was included. We excluded interventional studies, studies using AI‐based methods for IPFD quantification, studies with small total sample size (n < 20), and studies assessing LFC and IPFD using non‐MRI techniques (e.g., computed tomography, transabdominal ultrasonography). Conference abstracts and non‐original studies were also excluded. Two investigators (H.H.S. and S.A.) independently conducted the literature search after deduplication. Following screening of titles and abstracts to remove irrelevant studies, full‐text articles were assessed for eligibility. Any discrepancies were resolved through discussion with a third investigator (M.S.).

2.2. Data Abstraction and Risk of Bias Assessment

Using a standardised format, we extracted relevant study‐level data on study characteristics, study populations, and outcomes of interest. Where applicable, data reported as median (interquartile range or range) were converted to mean (standard deviation) using established methods [13]. We also extracted IPFD data from graphs using a digital plot extraction tool (WebPlotDigitizer, version 4.2). When only region‐specific pancreatic regions of interest (ROIs) were reported, whole‐pancreas IPFD was calculated as the mean of the available ROIs. Risk of bias in each included study was assessed using the Newcastle‐Ottawa Scale [14] by two independent investigators (S.A. and H.H.S.), with discrepancies resolved through discussion with a third investigator (M.S.).

2.3. Statistical Analysis

All analyses were conducted using R software (version 4.4.0). A p‐value of less than 0.05 defined statistical significance. A random‐effects meta‐analysis was performed to obtain pooled standardised mean differences (SMDs) in LFC between individuals with T2DM and non‐diabetic control groups in dual‐organ MRI studies, along with corresponding 95% confidence intervals (CIs). Effect sizes were aggregated using the restricted maximum likelihood method with the Hartung‐Knapp adjustment. Between‐study heterogeneity was assessed through the I ² statistics and between‐study variance (τ ²). To examine whether study‐level LFC differences varied according to mean IPFD levels in the T2DM group, exploratory subgroup analysis compared pooled SMDs in LFC between studies with low versus high mean IPFD levels (< 10% vs. ≥ 10%), based on previous evidence [15, 16]. Given the lack of a universal threshold for high IPFD, and consistent with the caution expressed in the 2026 Melbourne consensus of experts on IPFD [17], ancillary analyses were performed using alternative cut‐offs in 2% increments (e.g., < 6% vs. ≥ 6% up to < 12% vs. ≥ 12%). Additional subgroup analyses were performed according to geographic region (Asian vs. non‐Asian populations), MRI technique (two‐/three‐echo Dixon vs. multi‐echo Dixon), and risk of bias (low vs. moderate/high). Univariable meta‐regressions assessed study‐level associations between SMD in LFC and potential moderators, including SMD in IPFD and SMD in BMI. Multivariable meta‐regression was then used to test whether significant associations persisted after adjustment. Sensitivity analyses included restriction to proton density fat fraction (PDFF)‐based measurements of IPFD, with IPFD subgroup and meta‐regression analyses repeated in this subset, and leave‐one‐out analysis.

3. Results

3.1. Study Selection and Characteristics

The initial search identified 2837 records. After removal of duplicates and screening of titles and abstracts for irrelevant articles, 166 records were retained for full‐text review. Ultimately, 17 dual‐organ MRI studies (n = 2142) met the eligibility criteria (Figure 1). Table S2 summarises the study and patient characteristics. Most studies were cross‐sectional (n = 15), mainly conducted in Asian countries (n = 11). LFC and IPFD were most commonly assessed using multi‐echo Dixon (n = 12). T2DM definitions varied across studies, from formal diagnostic criteria to physician‐documented or patient‐reported diagnosis. Detailed MRI characteristics are presented in Table S3. The risk of bias assessment for all included studies is provided in Table S4.

FIGURE 1.

Association between SMD IPFD and SMD LFC across all included studies. Data are from univariable meta‐regression analysis. IPFD, intrapancreatic fat deposition; LFC, liver fat content; SMD, standardised mean difference.

3.2. Overall Liver Fat Excess in T2DM

The pooled analysis showed a statistically significant increase in LFC in individuals with T2DM compared with non‐diabetic controls (SMD 0.89, 95% CI 0.60–1.18; p < 0.0001), with high heterogeneity (I ² = 87.0%, τ ² = 0.27) (Figure S2). In the sensitivity analysis restricted to MRI‐PDFF studies, the pooled estimate remained directionally consistent with the main analysis (SMD 0.98, 95% CI 0.54–1.43; p < 0.001) and heterogeneity remained high (Figure S3). In the leave‐one‐out sensitivity analysis, no single study exerted a disproportionate influence on the overall effect estimate (Figure S4).

3.3. Liver Fat Excess in T2DM According to IPFD Level

When studies were stratified according to mean IPFD levels in the T2DM group, T2DM was associated with significantly higher LFC in both strata (Figure S5). In studies with low IPFD (< 10%), the pooled SMD in LFC was 0.62 (95% CI 0.41 to 0.83; I ² = 45.6%, τ ² = 0.04). In studies with high IPFD (≥ 10%), the corresponding pooled SMD was 1.20 (95% CI 0.53–1.87; I ² = 92.3%, τ ² = 0.47). In this exploratory subgroup analysis, liver fat excess was greater in studies with high mean IPFD in the T2DM group (p = 0.04). In leave‐one‐out analyses, the high IPFD stratum consistently showed numerically larger LFC differences than the low IPFD stratum, although the subgroup difference was not consistently statistically significant (p range: 0.012–0.13) (Table S5). When restricted to MRI‐PDFF studies, the difference between the < 10% and ≥ 10% IPFD strata was attenuated and was no longer statistically significant (p = 0.104) (Figure S6).

Ancillary analyses using alternative IPFD cut‐offs showed the same direction of effect, with numerically greater liver fat excess consistently observed in the high IPFD strata at 6%, 8%, and 12% thresholds; however, these subgroup differences did not reach the predefined level of statistical significance (Table 1).

3.4. Influence of Other Study‐Level Factors

No differences were found in pooled SMD when stratified by world region (p = 0.88), MRI technique (p = 0.32), and risk of bias (p = 0.31) (Table S6). In univariable meta‐regression, greater differences in IPFD were significantly associated with greater differences in LFC (β = 0.383, 95% CI 0.127–0.640; p = 0.006), explaining 41.0% of between‐study heterogeneity (Figure 1). Also, greater differences in BMI were significantly associated with greater differences in LFC (β = 0.288, 95% CI 0.159–0.416; p = 0.0003), explaining 63.6% of between‐study heterogeneity. In the multivariable model including both SMD in BMI and SMD in IPFD, the association between IPFD and LFC differences was attenuated and was no longer statistically significant (p = 0.596) (Table S7). In the PDFF‐restricted meta‐regression, IPFD differences remained significantly associated with LFC differences in univariable analysis but not after adjustment for BMI (Table S8).

4. Discussion

In this meta‐analysis of dual‐organ MRI studies, LFC was significantly greater in individuals with T2DM than in non‐diabetic controls. Specifically, it was approximately 0.9 standard deviations higher in the T2DM group, representing a large effect size and a substantial separation between the groups. Furthermore, a consistent pattern was observed in stratified analyses, with greater liver fat excess in studies with high compared with low mean IPFD. A statistically significant difference between these strata was observed using a 10% IPFD cut‐off in T2DM. Meta‐regression also showed that larger T2DM–control differences in IPFD were associated with larger differences in LFC. However, this association was attenuated and no longer statistically significant after adjusting for study‐level differences in BMI.

The present study complements our previous meta‐analysis quantifying excess IPFD in T2DM compared with non‐diabetic controls [12]. In contrast, the current synthesis focuses on LFC as the primary outcome in the subset of dual‐organ MRI studies, investigating whether the magnitude of liver fat excess in T2DM varies according to study‐level differences in IPFD. These findings are relevant because higher IPFD may reflect a broader multi‐organ ectopic fat phenotype in T2DM that also encompasses increased LFC [18, 19]. This perspective is supported by growing evidence that excessive IPFD, as a measurable morphological alteration of the pancreas, is associated with adverse pancreatic outcomes, broader metabolic derangements, and higher mortality [20, 21, 22]. Recent meta‐analyses also indicate that IPFD is potentially modifiable, with reductions reported following lifestyle interventions, glucose‐lowering medications, and bariatric surgery [23, 24, 25]. From a clinical perspective, combined assessment of liver and intrapancreatic fat may provide a more comprehensive characterisation of multi‐organ ectopic fat deposition. However, such assessment is not currently recommended for routine risk stratification in T2DM, and its incremental clinical value remains uncertain. Future studies using individual‐level data are needed to clarify the relationship between IPFD and LFC and to determine whether this association persists after accounting for BMI and other relevant factors.

The present study also contributes to the ongoing uncertainty regarding the optimal threshold for defining high IPFD [17, 18]. Although IPFD cut‐offs have been suggested [18, 26, 27], the Melbourne consensus of global experts on IPFD emphasised that establishing a universal threshold is premature given the limited evidence [18] in a field that has only recently adopted its first conceptual framework [17]. One challenge is the heterogeneity of approaches used to quantify IPFD, including emerging AI‐based methods. The accuracy of AI‐based approaches for measuring IPFD remains suboptimal [28], and such methods are therefore not routinely used in clinical practice. Accordingly, this systematic review deliberately excluded studies (including those derived from the UK Biobank) that used AI‐based methods for IPFD assessment. In this context, the present study provides new evidence by examining differences in LFC between individuals with T2DM and non‐diabetic controls across study‐level IPFD strata. A statistically significant subgroup difference was observed when studies were stratified using a 10% IPFD threshold. Although alternative thresholds showed a similar pattern, these subgroup differences did not reach statistical significance, potentially reflecting the limited number of studies within some strata. Therefore, the 10% IPFD threshold should be regarded as exploratory and hypothesis‐generating rather than as a validated clinical cut‐off. Further methodological and longitudinal studies will be required before an optimal IPFD threshold can be established.

This study has several limitations. First, because the analysis was restricted to dual‐organ MRI studies, the pooled LFC estimate should be interpreted as representative of this specific subset rather than as a comprehensive estimate of the overall LFC burden in T2DM. Second, substantial between‐study heterogeneity was observed in the primary analysis. Although exploratory analyses were performed to investigate potential sources of heterogeneity, residual heterogeneity remained, warranting cautious interpretation of the pooled estimates. Third, all analyses were based on aggregated study‐level data; therefore, the observed associations should not be interpreted at the individual level. Last, the diagnostic criteria for T2DM were not uniform across the included studies, and limited reporting precluded adjustment for potentially important clinical and adiposity‐related variables, including glycated haemoglobin, diabetes duration, background glucose‐lowering therapy, and visceral adipose tissue—all of which may have influenced the relationship between IPFD and LFC.

In conclusion, this meta‐analysis demonstrates that individuals with T2DM exhibit a large excess of LFC compared with non‐diabetic controls in dual‐organ MRI studies. Study‐level analyses further suggest that liver fat excess is greater in studies reporting higher IPFD. These findings support more in‐depth investigation of the interrelationship between IPFD and LFC in T2DM.

Author Contributions

Hyun Hee Sul: formal analysis, data curation, methodology, investigation, writing – review and editing. Saeed Ahmad: formal analysis, data curation, methodology, investigation, writing – review and editing. Pojsakorn Danpanichkul: investigation, writing – review and editing. Matheus Souza: formal analysis, data curation, methodology, investigation, writing – original draft, writing – review and editing, supervision, conceptualization. Maxim S. Petrov: investigation, writing – review and editing, supervision, conceptualization.

Funding

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: PRISMA checklist.

Table S2: Main characteristics of the populations included in the analyses.

Table S3: MRI characteristics of the included studies.

Table S4: Risk of bias assessment of the included studies.

Table S5: Leave‐one‐out sensitivity analyses for the IPFD subgroup comparison.

Table S6: Subgroup analyses of pooled liver fat content differences between T2DM and non‐diabetic controls.

Table S7: Study‐level factors associated with liver fat content differences between individuals with Type 2 diabetes and non‐diabetic controls.

Table S8: Sensitivity analysis for meta‐regressions restricted to MRI‐PDFF studies.

Figure S1: PRISMA flow diagram. PRISMA, Preferred Reporting Items for Systematic Reviews and Meta‐Analyses.

Figure S2: Differences in liver fat content between individuals with Type 2 diabetes mellitus and non‐diabetic controls.

Figure S3: Sensitivity analysis restricted to MRI‐PDFF studies.

Figure S4: Leave‐one‐out sensitivity analysis.

Figure S5: Subgroup analysis of differences in liver fat content between individuals with Type 2 diabetes mellitus and non‐diabetic controls, stratified by mean IPFD in the T2DM group (< 10% vs. ≥ 10%).

Figure S6: Sensitivity analysis for subgroup analysis restricted to MRI‐PDFF studies.

Sul H. H., Ahmad S., Danpanichkul P., Souza M., and Petrov M. S., “Excess Liver Fat in Type 2 Diabetes Stratified by Intrapancreatic Fat Deposition: A Meta‐Analysis of Dual‐Organ MRI Studies,” Liver International 46, no. 11 (2026): e70934, 10.1111/liv.70934.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.