Burden of Intrapancreatic Fat Deposition in Type 2 Diabetes and the Role of Obesity: A Systematic Review and Meta-Analysis Authors: Hyun Hee Sul, Saeed Ahmad, Pojsakorn Danpanichkul, Alessandro Mantovani, Matheus Souza, Maxim S. Petrov Published: 2026 ABSTRACT BackgroundGrowing evidence supports a role for high intrapancreatic fat deposition (IPFD) in the pathogenesis of type 2 diabetes mellitus (T2DM); however, the magnitude of this association and the extent to which it is influenced by body mass index (BMI), liver fat content (LFC), and age remain uncertain.ObjectiveTo quantitatively evaluate IPFD measured by magnetic resonance imaging (MRI) in individuals with T2DM and to investigate study-level factors contributing to between-study variability.MethodsWe systematically searched PubMed and Embase for observational studies comparing MRI-measured IPFD in individuals with T2DM versus non-diabetic controls, excluding studies in which IPFD quantification was performed using AI-based models. Pooled standardized mean differences (SMDs) were estimated using a restricted maximum likelihood approach with Hartung-Knapp adjustment. Heterogeneity was explored through subgroup, meta-regression, and sensitivity analyses.ResultsThirty studies (n=3,980) were included. Individuals with T2DM had significantly higher IPFD than non-diabetic controls (SMD 1.13, 95% CI 0.74 to 1.51), with no significant differences when stratified by pancreatic region, MRI technique, geographic region, sample size, and risk of bias. In multivariable models, BMI remained the only consistent moderating covariate, and the excess IPFD in T2DM persisted after adjusting for BMI, LFC, and age (intercept=0.797, p=0.010, false discovery rate-adjusted p=0.018). Advanced heterogeneity exploration identified no structural outliers, and the pooled effect size remained consistent across multiple sensitivity analyses.ConclusionT2DM is associated with substantially higher MRI-quantified IPFD, independent of study-level differences in BMI, LFC, and age. Although causality cannot be inferred, these findings support the hypothesis that IPFD may be involved in the underlying mechanism of T2DM. They also suggest that high IPFD could represent a metabolic hallmark, with potential implications for risk stratification and therapeutic targeting. Graphical Abstract Supplementary InformationThe online version contains supplementary material available at 10.1007/s13679-026-00718-3. Introduction Intrapancreatic fat deposition (IPFD) has emerged as an important component of ectopic fat accumulation and a potential contributor to systemic metabolic dysfunction, extending beyond the well-known sites of the liver and visceral adipose tissue [1]. Over the past decade, IPFD has attracted increasing scientific attention, driven largely by advances in imaging-based characterization of multi-organ ectopic fat burden [2]. While the pancreas normally contains a small amount of fat, excessive IPFD has been increasingly associated with β-cell dysfunction, insulin resistance, and alterations in the islet microenvironment [3–6]. Further, longitudinal cohort studies indicate that excessive IPFD is independently associated with a higher risk of incident type 2 diabetes mellitus (T2DM) [7–10]. A key challenge in studying IPFD is that its biological burden in T2DM remains largely unknown. Specifically, the extent to which IPFD increases during the natural course of T2DM, in the absence of interventions, has not been clearly determined. Although several meta-analyses have explored associations between IPFD and T2DM, they have primarily focused on risk ratios or odds ratios, which depend heavily on the imaging modality and the thresholds used to define excessive IPFD [3, 11–14]. Techniques such as transabdominal and endoscopic ultrasound are highly operator-dependent and qualitative, making them unreliable for establishing excessive IPFD. Computed tomography estimates IPFD only indirectly using X-ray attenuation values [15, 16], whereas magnetic resonance spectroscopy (MRS) quantifies IPFD directly but is susceptible to voxel placement errors and partial volume effects, including contamination from surrounding visceral fat [1, 5, 17]. As such, both techniques remain suboptimal for accurate IPFD measurement. In contrast, chemical-shift-encoded magnetic resonance imaging (MRI), which enables quantitative assessment of proton density fat fraction (PDFF), is a well validated and widely used non-invasive method for the quantification of IPFD [18]. It permits evaluation of the extent to which IPFD differs between individuals with T2DM and non-diabetic controls. For understanding the role of IPFD in T2DM, the magnitude of such differences provides more meaningful insight than prevalence above arbitrary thresholds. These quantitative insights could also guide the development and targeting of interventions, with implications for improved care of individuals with T2DM [19–21]. Another important factor that remains underexplored in the relationship between IPFD and T2DM is obesity. Although body mass index (BMI) is often assumed to correlate directly with IPFD, no systematic studies have rigorously assessed the role of obesity in modulating this association. Other factors potentially influencing the relationship between IPFD and T2DM have also not been systematically investigated. The primary aim of this study was to conduct a systematic literature review and meta-analysis to quantify the excess IPFD in individuals with T2DM compared with non-diabetic controls. The secondary aim was to examine the roles of obesity and other potential contributing factors in this relationship. Methods Data Sources and Searches This systematic review followed the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) and Meta-analysis Of Observational Studies in Epidemiology (MOOSE) guidelines [22, 23]. The review protocol was prospectively registered in the International Prospective Register of Systematic Reviews (PROSPERO) database (CRD420251208380). We systematically searched two major electronic databases (PubMed and Embase) from inception till November 15, 2025. The predefined search strategy is detailed in the Supplementary Methods. In addition, reference lists of included studies and relevant reviews were manually screened, and the authors’ personal libraries were consulted to identify any additional eligible studies. Eligibility Criteria After deduplication, two authors (H.H.S. and S.A.) independently screened study titles and abstracts. Full-text articles were then assessed for eligibility according to the predefined inclusion criteria. Any discrepancies were resolved by consultation with a third author (M.S.). Studies were eligible for inclusion if they met the following criteria: (i) observational design and inclusion of patients diagnosed with T2DM; (ii) quantitative assessment of IPFD using chemical shift-encoded MRI (e.g., Dixon, PDFF); (iii) provision of sufficient data to calculate differences in IPFD between patients with T2DM and non-diabetic control participants. The diagnosis of T2DM was accepted as reported in the primary studies. Control participants were defined as individuals without diabetes or prediabetes. Interventional studies (randomized or non-randomized) were excluded, as these predominantly assess changes in IPFD rather than baseline burden and may therefore confound evaluation of the natural IPFD burden [19–21, 24, 25]. Studies using AI–based methods for quantification of IPFD were excluded, as these modern approaches currently demonstrate suboptimal accuracy compared with manual quantification [26]. Studies that quantified IPFD using H1-MRS were also excluded [17]. Only original articles were considered for inclusion. Studies that were conference abstracts, focused on highly selected populations (e.g., pediatric, gestational diabetes mellitus, individuals after acute pancreatitis), non-original works (e.g., guidelines, editorials, and reviews), or had a small total sample size (< 20 participants) were excluded. When multiple reports from the same cohort were identified, only the most comprehensive publication was included to avoid data duplication. Data Extraction and Risk of Bias Assessment The following study-level data were extracted using a standardized form: (i) study characteristics: first author, publication year, location, study year, setting/population, MRI characteristics, and T2DM diagnosis; (ii) population characteristics: sample size, age, sex, BMI, liver fat content [LFC], HbA1c levels; and (iii) outcome of interest. Data transformations were performed according to the guidelines in the Cochrane Handbook and the recommended formulas, when applicable [27]. Consistent with prior MRI studies, whole-pancreas IPFD was estimated using the mean of the three pancreatic regions of interest (ROIs: head, body, and tail) when volume-weighted measures were unavailable [28]. If studies reported IPFD measures for only a single ROI, these data were used exclusively in the corresponding subgroup analysis [29]. When data were only available as graphs or bar charts [2], values were manually extracted using a free web-based digitizing tool (WebPlotDigitizer v4.2), following the recommended methods for standardized quantitative synthesis [30]. Two authors (S.A. and H.H.S.) independently assessed the risk of bias in each study using the Newcastle-Ottawa Scale [31]. Disagreements were resolved through evaluation by a third author (M.S.). Publication bias was evaluated using visual inspection of funnel plot asymmetry. Outcomes of Interest The primary outcome was the standardized mean difference (SMD) in IPFD between patients with T2DM and non-diabetic controls, as quantified by chemical shift-encoded MRI. To improve the clinical interpretability of IPFD differences, we also estimated mean differences (MDs) in the overall analysis. Data Synthesis and Statistical Analysis All statistical analyses were performed using R software (v4.4.0; “meta” and “metafor” packages). A two-tailed p-value < 0.05 was considered statistically significant. Pooled SMDs/MDs and their 95% confidence intervals (CIs) were estimated using the restricted maximum likelihood (REML) method and the Hartung-Knapp adjustment to provide robust variance estimates [32]. Heterogeneity was evaluated using the between-study variance (τ2) and the I² statistic. To investigate potential sources of heterogeneity, subgroup analyses were performed according to specific pancreatic region assessed, MRI-based technique, world location, sample size, and risk of bias. We then performed univariable meta-regressions, using study-level SMDs for age, BMI and LFC, as well as proportion of women as potential moderators of the associations with SMD IPFD. Given the expected interdependence between IPFD and metabolic traits, we further performed multivariable meta-regressions using prespecified models (Model 1: SMD BMI + SMD Age; Model 2: SMD BMI + SMD LFC; Model 3: SMD BMI + SMD Age + SMD LFC). Importantly, women proportion was not included in the multivariable models because, unlike the SMD-based moderators, it reflected study composition rather than a between-group contrast. The proportion of heterogeneity explained by each meta-regression model was summarized using the R2 analog. To account for multiple testing, p-values for moderator coefficients were adjusted using the Benjamini-Hochberg false discovery rate (FDR) procedure [33]. Because the intercept of the most adjusted model was also used to assess whether excess IPFD persisted after covariate adjustment, we additionally applied FDR correction including the model intercepts. FDR-adjusted p-values (i.e., q-values) < 0.05 were considered statistically significant. To characterize the structure of heterogeneity, we conducted Graphic Display of Study Heterogeneity (GOSH) analyses using 50,000 random subsets of studies under the REML estimator [34]. Density-based clustering (DBSCAN) was applied to the GOSH space (pooled SMD × τ2) to identify homogeneous clusters and potential outlier substructures. We conducted a sensitivity analysis restricting the analysis to studies with an average non-obese population. Obesity status was determined using the standard World Health Organization cutoff for non-Asian populations (BMI ≥ 30 kg/m²) and an adjusted cutoff for Asian populations (BMI ≥ 28 kg/m²). Leave-one-out sensitivity analysis was also performed to evaluate whether any single study disproportionately affected the pooled SMD. Post hoc analysis was performed by restricting the meta-analysis to the studies within the DBSCAN-identified homogeneous cluster. Results Search Results A comprehensive literature search yielded 2,837 records, of which 408 duplicates were removed. After title and abstract screening, 2,429 records were assessed and 166 studies were retrieved for full-text eligibility evaluation. We also identified 2 additional eligible studies via non-database sources and manual searching [35, 36]. Ultimately, 30 studies (n = 3,980 individuals) were included in the meta-analysis [2, 28, 29, 35–61]. Fig. 1 presents the PRISMA diagram summarizing the study selection processes.Fig. 1PRISMA flow diagram. PRISMA, Preferred Reporting Items for Systematic reviews and Meta-Analyses PRISMA flow diagram. PRISMA, Preferred Reporting Items for Systematic reviews and Meta-Analyses Characteristics of Included Studies The main characteristics of the included studies are presented in Table 1. Most studies were cross-sectional in design (90%) [28, 29, 36–56, 58–61] and conducted in clinic or hospital-based settings (90%) [2, 28, 29, 35, 36, 38–46, 48–54, 56–61] . Geographically, the majority were performed in Asia (60%) [29, 36, 40–42, 45, 46, 48–54, 56, 58, 60, 61] . MRI characteristics are detailed in Table S1. Most studies reported whole-pancreas IPFD measurement (86.7%) [2, 35–57, 59, 60], with a subset also providing region-specific measurements [2, 40, 42, 44, 46, 49, 56]. Three studies reported only region-specific values for the head, body, and tail [28, 58, 61], whereas one study reported data exclusively for the pancreatic head [29]. Baseline patient characteristics stratified by T2DM status are reported in Table S2. Twelve studies were judged to have a low risk of bias based on the Newcastle–Ottawa Scale [2, 28, 43, 47, 49–51, 54, 58–61], and Table S3 summarizes the risk of bias assessments for all included studies.Table 1Main characteristics of the included studiesFirst author (year)CountryStudy designMRI techniquePancreatic locationSample size, n#T2DM, %Ma et al. [29] (2014)ChinaCross-sectionalMulti-echo Dixon (6-echo)Regions only (head)3470.6Kühn et al. [37] (2015)GermanyCross-sectionalThree-echo DixonWhole pancreas only8108.6Idilman et al. [38] (2015)TurkeyCross-sectionalMulti-echo Dixon (6-echo)Whole pancreas only4112.2Macauley et al. [39] (2015)UKCross-sectionalThree-echo DixonWhole pancreas only5574.5Chai et al. [40](2016)ChinaCross-sectionalTwo-echo DixonWhole pancreas and regions (head, body, tail)10070.0Heber et al. [2] (2017)GermanyCase-controlMulti-echo Dixon (6-echo)Whole pancreas and regions (head, body, tail)29018.3Lu et al. [41] (2019)ChinaCross-sectionalThree-echo DixonWhole pancreas only11369.0Wang et al. [42] (2019)ChinaCross-sectionalMulti-echo Dixon (6-echo)Whole pancreas and regions (head, body, tail)3148.4Tirkes et al. [43] (2019)USACross-sectionalTwo-echo DixonWhole pancreas only11813.6Nadarajah et al. [28] (2020)USACross-sectionalTwo-echo DixonRegions only(head/body/tail)19523.1Sarma et al. [44] (2020)USACross-sectionalMulti-echo Dixon (6-echo)Whole pancreas and regions (head, body/tail)2751.9Li et al. [45] (2021)ChinaCross-sectionalThree-echo DixonWhole pancreas only6011.7Zheng et al. [46] (2022)ChinaCross-sectionalTwo-echo DixonWhole pancreas and regions (head, body, tail)15775.2Waddell et al. [47] (2022)UKCross-sectionalMulti-echo Dixon (10-echo)Whole pancreas only26649.2Wen et al. [48] (2022)ChinaCross-sectionalMulti-echo Dixon (6-echo)Whole pancreas only3447.1Yu et al. [49] (2023)ChinaCross-sectionalMulti-echo Dixon (6-echo)Whole pancreas and regions (head, body, tail)8234.1Yi et al. [50] (2023)ChinaCross-sectionalMulti-echo Dixon (6-echo)Whole pancreas only9549.5Cao et al. [51] (2023)ChinaCross-sectionalMulti-echo Dixon (6-echo)Whole pancreas only9243.5Yasokawa et al. [52] (2023)JapanCross-sectionalMulti-echo Dixon (6-echo)Whole pancreas only5541.8Ting et al. [53] (2023)ChinaCross-sectionalThree-echo DixonWhole pancreas only10070.0Wang et al. [36] (2023)ChinaCross-sectionalTwo-echo DixonWhole pancreas only10080.0An et al. [54] (2024)ChinaCross-sectionalMulti-echo Dixon*Whole pancreas only34119.9Diamond et al. [55] (2024)UKCross-sectionalMulti-echo Dixon*Whole pancreas only20355.2Qu et al. [56] (2025)ChinaCross-sectionalMulti-echo Dixon*Whole pancreas and regions (head, body, tail)8050.0Elsayed et al. [57] (2025)EgyptCase-controlMulti-echo Dixon (6-echo)Whole pancreas only3050.0Yuan et al. [58] (2025)ChinaCross-sectionalMulti-echo Dixon (6-echo)Regions only(head/body/tail)11867.8Akhan et al. [59] (2025)TurkeyCross-sectionalTwo-echo DixonWhole pancreas only8664.0Nie et al. [60] (2025)ChinaCross-sectionalMulti-echo Dixon*Whole pancreas only9647.9Nielsen et al. [35] (2025)DenmarkCase-controlMulti-echo Dixon*Whole pancreas only2352.2Zhu et al. [61] (2025)ChinaCross-sectionalMulti-echo Dixon*Regions only(head/body/tail)14850.0* In these instances, the exact number of echoes was not clearly reported# In studies where a subset of participants with prediabetes was identified, these individuals were excluded from the analysisAbbreviations: MRI magnetic resonance imaging, T2DM type 2 diabetes mellitus Table 1Main characteristics of the included studiesFirst author (year)CountryStudy designMRI techniquePancreatic locationSample size, n#T2DM, %Ma et al. [29] (2014)ChinaCross-sectionalMulti-echo Dixon (6-echo)Regions only (head)3470.6Kühn et al. [37] (2015)GermanyCross-sectionalThree-echo DixonWhole pancreas only8108.6Idilman et al. [38] (2015)TurkeyCross-sectionalMulti-echo Dixon (6-echo)Whole pancreas only4112.2Macauley et al. [39] (2015)UKCross-sectionalThree-echo DixonWhole pancreas only5574.5Chai et al. [40](2016)ChinaCross-sectionalTwo-echo DixonWhole pancreas and regions (head, body, tail)10070.0Heber et al. [2] (2017)GermanyCase-controlMulti-echo Dixon (6-echo)Whole pancreas and regions (head, body, tail)29018.3Lu et al. [41] (2019)ChinaCross-sectionalThree-echo DixonWhole pancreas only11369.0Wang et al. [42] (2019)ChinaCross-sectionalMulti-echo Dixon (6-echo)Whole pancreas and regions (head, body, tail)3148.4Tirkes et al. [43] (2019)USACross-sectionalTwo-echo DixonWhole pancreas only11813.6Nadarajah et al. [28] (2020)USACross-sectionalTwo-echo DixonRegions only(head/body/tail)19523.1Sarma et al. [44] (2020)USACross-sectionalMulti-echo Dixon (6-echo)Whole pancreas and regions (head, body/tail)2751.9Li et al. [45] (2021)ChinaCross-sectionalThree-echo DixonWhole pancreas only6011.7Zheng et al. [46] (2022)ChinaCross-sectionalTwo-echo DixonWhole pancreas and regions (head, body, tail)15775.2Waddell et al. [47] (2022)UKCross-sectionalMulti-echo Dixon (10-echo)Whole pancreas only26649.2Wen et al. [48] (2022)ChinaCross-sectionalMulti-echo Dixon (6-echo)Whole pancreas only3447.1Yu et al. [49] (2023)ChinaCross-sectionalMulti-echo Dixon (6-echo)Whole pancreas and regions (head, body, tail)8234.1Yi et al. [50] (2023)ChinaCross-sectionalMulti-echo Dixon (6-echo)Whole pancreas only9549.5Cao et al. [51] (2023)ChinaCross-sectionalMulti-echo Dixon (6-echo)Whole pancreas only9243.5Yasokawa et al. [52] (2023)JapanCross-sectionalMulti-echo Dixon (6-echo)Whole pancreas only5541.8Ting et al. [53] (2023)ChinaCross-sectionalThree-echo DixonWhole pancreas only10070.0Wang et al. [36] (2023)ChinaCross-sectionalTwo-echo DixonWhole pancreas only10080.0An et al. [54] (2024)ChinaCross-sectionalMulti-echo Dixon*Whole pancreas only34119.9Diamond et al. [55] (2024)UKCross-sectionalMulti-echo Dixon*Whole pancreas only20355.2Qu et al. [56] (2025)ChinaCross-sectionalMulti-echo Dixon*Whole pancreas and regions (head, body, tail)8050.0Elsayed et al. [57] (2025)EgyptCase-controlMulti-echo Dixon (6-echo)Whole pancreas only3050.0Yuan et al. [58] (2025)ChinaCross-sectionalMulti-echo Dixon (6-echo)Regions only(head/body/tail)11867.8Akhan et al. [59] (2025)TurkeyCross-sectionalTwo-echo DixonWhole pancreas only8664.0Nie et al. [60] (2025)ChinaCross-sectionalMulti-echo Dixon*Whole pancreas only9647.9Nielsen et al. [35] (2025)DenmarkCase-controlMulti-echo Dixon*Whole pancreas only2352.2Zhu et al. [61] (2025)ChinaCross-sectionalMulti-echo Dixon*Regions only(head/body/tail)14850.0* In these instances, the exact number of echoes was not clearly reported# In studies where a subset of participants with prediabetes was identified, these individuals were excluded from the analysisAbbreviations: MRI magnetic resonance imaging, T2DM type 2 diabetes mellitus Main characteristics of the included studies Regions only (head/body/tail) Regions only (head/body/tail) Regions only (head/body/tail) * In these instances, the exact number of echoes was not clearly reported # In studies where a subset of participants with prediabetes was identified, these individuals were excluded from the analysis Abbreviations: MRI magnetic resonance imaging, T2DM type 2 diabetes mellitus Burden of Intrapancreatic Fat Deposition Figure 2 shows the distribution of effect estimates across observational studies evaluating MRI-based IPFD in participants with T2DM. Overall, individuals with T2DM had significantly higher IPFD compared with non-diabetic controls (SMD 1.13, 95% CI 0.74 to 1.51; p < 0.0001), with substantial between-study heterogeneity (I²=89.8%, τ2 = 0.81). When expressed as absolute IPFD change, the pooled MD was 3.86% (95% CI 2.76 to 4.96, I²=94%, τ2 = 5.70; p < 0.0001; Fig. 3).Fig. 2Differences in IPFD between type 2 diabetes and controls across all included studies. IPFD, intrapancreatic fat depositionFig. 3Absolute differences in IPFD between type 2 diabetes and controls across all included studies. IPFD, intrapancreatic fat deposition Differences in IPFD between type 2 diabetes and controls across all included studies. IPFD, intrapancreatic fat deposition Absolute differences in IPFD between type 2 diabetes and controls across all included studies. IPFD, intrapancreatic fat deposition GOSH exploration revealed a continuous heterogeneity landscape, with pooled SMDs spanning a wide range of τ2 values and no discrete “islands” suggestive of structural outliers (Fig. 4). Application of DBSCAN to the GOSH space identified one dominant cluster (cluster 1: 49,279/50,000 subsets; 98.6%), a small proportion of subsets classified as noise (cluster 0: 694/50,000; 1.4%), and very small peripheral clusters (clusters 2: 27/50,000; <0.1%). The dominant cluster exhibited the lowest mean τ² among non-noise clusters (mean τ² = 0.43) compared with the peripheral cluster (mean τ² = 1.16) and noise (mean τ² = 0.96), indicating a robust “core” region in the heterogeneity landscape. All studies were consistently represented within this main cluster (~ 49–51% across studies). Post hoc analysis restricted to studies within the dominant cluster yielded identical pooled estimates to the full model (SMD 1.13, 95% CI 0.75 to 1.51; I² = 94.6%, τ² = 0.83; p < 0.0001). A sensitivity analysis using the leave-one-out approach did not significantly affect the overall estimate (Figure S1).Fig. 4Advanced investigation of between-study heterogeneity. Graphic Display of Study Heterogeneity (GOSH) plot derived from 50,000 random subsets of the included studies, showing the relationship between pooled effect size (SMD) and between-study variance (τ2) estimated under random-effects modeling. Points are colored by DBSCAN cluster membership, which identified one dominant homogeneous core cluster (Cluster 1: 49,279/50,000 subsets) and a small proportion of subsets classified as noise (Cluster 0: 694/50,000), with a very small peripheral cluster (Clusters 2: 27/50,000) Advanced investigation of between-study heterogeneity. Graphic Display of Study Heterogeneity (GOSH) plot derived from 50,000 random subsets of the included studies, showing the relationship between pooled effect size (SMD) and between-study variance (τ2) estimated under random-effects modeling. Points are colored by DBSCAN cluster membership, which identified one dominant homogeneous core cluster (Cluster 1: 49,279/50,000 subsets) and a small proportion of subsets classified as noise (Cluster 0: 694/50,000), with a very small peripheral cluster (Clusters 2: 27/50,000) Influence of Obesity A sensitivity analysis restricted to studies of average non-obese populations demonstrated that individuals with T2DM had significantly higher IPFD compared with controls (SMD 1.27, 95% CI 0.55 to 1.99; p = 0.0025; Fig. 5), with substantial between-study heterogeneity (I²=93.9%, τ2 = 1.15). In a univariable meta-regression, the SMD for BMI statistically significant (coefficient = 0.473, 95% CI 0.284 to 0.662; p < 0.0001) and explained 64.9% of the between-study heterogeneity. In multivariable meta-regressions, the SMD for BMI persisted statistically significant when the models included LFC (coefficient = 0.568, 95% CI 0.272 to 0.863; p = 0.001) and included LFC together with age (coefficient = 0.583, 95% CI 0.306 to 0.860; p = 0.0006). These models explained 79.89% and 80.44% of the between-study heterogeneity, respectively (Table 2). After FDR correction, SMD BMI coefficients remained statistically significant across all models. The proportion of women and SMD age were not statistically significant after correction, whereas SMD LFC was significant only in univariable analysis and did not remain significant in models including SMD BMI. When the FDR correction was extended to include model intercepts, the intercept of the most adjusted model remained significant (q = 0.018) (Table 2).Fig. 5Differences in IPFD between individuals with type 2 diabetes and controls in average non-obese populations. IPFD, intrapancreatic fat depositionTable 2Factors influencing differences in IPFD between individuals with type 2 diabetes and controlskR 2Estimate95% confidence intervalP-valueQ-value*Q-value#Univariable meta-regressionWomen, %223.29%Intercept1.9110.556 to 3.2660.008-0.017Women, %−0.019−0.048 to 0.0100.1910.3180.274SMD BMI2164.90%Intercept0.6640.374 to 0.9540.0001-< 0.001SMD BMI0.4730.284 to 0.662< 0.0001< 0.001< 0.001SMD Age250%Intercept0.9630.550 to 1.375< 0.0001-< 0.001SMD Age0.198−0.461 to 0.8560.5410.6240.592SMD LFC1740.78%Intercept0.247−0.494 to 0.9880.489-0.591SMD LFC1.0410.328 to 1.7540.0070.0180.017Multivariable meta-regressionSMD BMI and SMD Age2161.74%Intercept0.6100.263 to 0.9570.002-0.005SMD BMI0.4740.281 to 0.666< 0.0001< 0.001< 0.001SMD Age0.151−0.343 to 0.6450.5290.6240.592SMD BMI and SMD LFC1679.89%Intercept0.8730.279 to 1.4680.007-0.017SMD BMI0.5680.272 to 0.8630.00110.0030.004SMD LFC−0.199−0.987 to 0.5880.5940.6360.621SMD BMI, SMD Age, and SMD LFC1680.44%Intercept0.7970.234 to 1.3600.010-0.018SMD BMI0.5830.306 to 0.8600.00060.0030.003SMD Age0.395−0.096 to 0.8870.1050.1970.161SMD LFC−0.271−1.012 to 0.4710.4420.6240.591* FDR-adjusted p-values for moderator coefficients only# FDR-adjusted p-values from a broader correction including model interceptsAbbreviations: R2: proportion of between-study heterogeneity explained by the moderator; SMD standardized mean difference, BMI body mass index, LFC liver fat content, FDR false discovery rate Differences in IPFD between individuals with type 2 diabetes and controls in average non-obese populations. IPFD, intrapancreatic fat deposition Table 2Factors influencing differences in IPFD between individuals with type 2 diabetes and controlskR 2Estimate95% confidence intervalP-valueQ-value*Q-value#Univariable meta-regressionWomen, %223.29%Intercept1.9110.556 to 3.2660.008-0.017Women, %−0.019−0.048 to 0.0100.1910.3180.274SMD BMI2164.90%Intercept0.6640.374 to 0.9540.0001-< 0.001SMD BMI0.4730.284 to 0.662< 0.0001< 0.001< 0.001SMD Age250%Intercept0.9630.550 to 1.375< 0.0001-< 0.001SMD Age0.198−0.461 to 0.8560.5410.6240.592SMD LFC1740.78%Intercept0.247−0.494 to 0.9880.489-0.591SMD LFC1.0410.328 to 1.7540.0070.0180.017Multivariable meta-regressionSMD BMI and SMD Age2161.74%Intercept0.6100.263 to 0.9570.002-0.005SMD BMI0.4740.281 to 0.666< 0.0001< 0.001< 0.001SMD Age0.151−0.343 to 0.6450.5290.6240.592SMD BMI and SMD LFC1679.89%Intercept0.8730.279 to 1.4680.007-0.017SMD BMI0.5680.272 to 0.8630.00110.0030.004SMD LFC−0.199−0.987 to 0.5880.5940.6360.621SMD BMI, SMD Age, and SMD LFC1680.44%Intercept0.7970.234 to 1.3600.010-0.018SMD BMI0.5830.306 to 0.8600.00060.0030.003SMD Age0.395−0.096 to 0.8870.1050.1970.161SMD LFC−0.271−1.012 to 0.4710.4420.6240.591* FDR-adjusted p-values for moderator coefficients only# FDR-adjusted p-values from a broader correction including model interceptsAbbreviations: R2: proportion of between-study heterogeneity explained by the moderator; SMD standardized mean difference, BMI body mass index, LFC liver fat content, FDR false discovery rate Factors influencing differences in IPFD between individuals with type 2 diabetes and controls * FDR-adjusted p-values for moderator coefficients only # FDR-adjusted p-values from a broader correction including model intercepts Abbreviations: R2: proportion of between-study heterogeneity explained by the moderator; SMD standardized mean difference, BMI body mass index, LFC liver fat content, FDR false discovery rate Influence of Other Factors In subgroup analyses, stratification by region of the pancreas (p = 0.99), MRI technique (p = 0.85), world location (p = 0.08), sample size (p = 0.68), or risk of bias (p = 0.94) did not significantly affect the differences in IPFD between individuals with T2DM and non-diabetic controls (Table 3). The funnel plot was relatively asymmetric on visual inspection (Figure S2).Table 3Pooled IPFD differences across subgroupsParameterStudies, nPatients, nSMD95% confidence intervalI², %τ2P-valueBy region of the pancreas Head1112621.240.25 to 2.2393.62.020.99 Body912011.160.25 to 2.0892.91.30 Tail912011.200.30 to 2.11931.28By MRI technique Two-/three-echo Dixon1118941.080.28 to 1.8889.31.110.85 Multi-echo Dixon1820521.160.70 to 1.63900.74By world location Asian1718021.360.75 to 1.9789.71.200.08 Non-Asian1221440.770.38 to 1.1683.90.24By sample size N < 100158671.210.62 to 1.79790.810.68 N ≥ 1001430591.050.48 to 1.6293.60.88By risk of bias Low1219271.140.49 to 1.7993.10.940.94 Moderate/high1720191.110.59 to 1.64860.77Abbreviations: SMD standardized mean difference, MRI magnetic resonance imaging Table 3Pooled IPFD differences across subgroupsParameterStudies, nPatients, nSMD95% confidence intervalI², %τ2P-valueBy region of the pancreas Head1112621.240.25 to 2.2393.62.020.99 Body912011.160.25 to 2.0892.91.30 Tail912011.200.30 to 2.11931.28By MRI technique Two-/three-echo Dixon1118941.080.28 to 1.8889.31.110.85 Multi-echo Dixon1820521.160.70 to 1.63900.74By world location Asian1718021.360.75 to 1.9789.71.200.08 Non-Asian1221440.770.38 to 1.1683.90.24By sample size N < 100158671.210.62 to 1.79790.810.68 N ≥ 1001430591.050.48 to 1.6293.60.88By risk of bias Low1219271.140.49 to 1.7993.10.940.94 Moderate/high1720191.110.59 to 1.64860.77Abbreviations: SMD standardized mean difference, MRI magnetic resonance imaging Pooled IPFD differences across subgroups Abbreviations: SMD standardized mean difference, MRI magnetic resonance imaging Discussion In this systematic review and meta-analysis of 30 observational studies encompassing 3,980 individuals, MRI-measured IPFD was substantially higher in individuals with T2DM compared with non-diabetic controls (SMD 1.13). Subgroup analyses revealed no significant differences in the magnitude of this association by world region, MRI techniques, pancreatic region, sample size, and risk of bias. Restricting the analysis to average non-obese populations did not attenuate the effect (SMD 1.27). Despite significant heterogeneity, multiple sensitivity analyses confirmed the robustness of the observed association. In univariable meta-regressions, SMD BMI was significantly associated with increased IPFD in T2DM. Multivariable models suggested that SMD BMI may act as an independent contributor to IPFD changes, whereas SMD LFC and SMD age did not significantly modify the association. Notably, even after adjustment for all these study-level variables, a considerable residual SMD IPFD persisted (intercept = 0.797, p = 0.010, q = 0.018). Our findings are consistent with the growing body of evidence highlighting the role of IPFD in T2DM pathophysiology. The PANDORA (PANcreatic Disease Originating from intRa-pancreatic fAt) hypothesis postulates a causative role of high IPFD in the development of T2DM (as well as major exocrine pancreatic diseases) [62]. This hypothesis is strongly supported by earlier longitudinal evidence [7–9, 63, 64], including evidence from lean individuals [65]. The robust association observed between T2DM and increased IPFD aligns with this hypothesis and is further supported by experimental and human data linking IPFD to β-cell dysfunction and insulin resistance, primarily mediated via lipotoxicity [5, 11, 48, 66]. Orthogonal evidence also shows that individuals with high genetic predisposition to insulin resistance have significantly higher IPFD [67]. Besides, the observed absolute increase in IPFD of 3.86% in individuals with T2DM is substantial relative to previously proposed normal thresholds [68, 69]. Our analyses demonstrated that the increase in IPFD among individuals with T2DM (SMD 1.13) persisted after accounting for LFC, age, and BMI. Notably, average non-obese populations with T2DM still exhibited a substantial effect (SMD 1.27) compared with controls. This finding aligns with a 2025 whole-organ quantitative histological analysis of adult donor pancreases, which reported higher adipocyte infiltration in both intra- and interlobular pancreatic compartments in individuals with T2DM versus non-diabetic controls [70]. The independence from obesity suggests that IPFD is not merely a bystander of general adiposity, but may arise through alternative mechanisms of fat accumulation in the pancreas, including lipid droplets within acinar cells and islets, adipose replacement following acinar cell apoptosis, and acinar-to-adipocyte transdifferentiation [1]. The observed gradient of IPFD from controls to T2DM reflects a continuum, reinforcing the notion that IPFD could serve as an early marker of metabolic dysfunction along the trajectory of T2DM (independent of the obesity status). This study has several important strengths. Building on previous individual studies of MRI-measured IPFD in T2DM [9], this study is the first meta-analysis to quantify the magnitude of the association between T2DM and IPFD using MRI exclusively. Our meta-analysis includes a large sample of MRI-based IPFD measurements from diverse populations across Asia, Europe, and North America. By applying rigorous inclusion criteria centered on MRI-based measurements performed by researchers (studies using AI-based measurements were ineligible [26]) techniques and excluding MRS studies, we ensured greater consistency and accuracy in the IPFD measurements analyzed. Heterogeneity was comprehensively evaluated using multiple approaches, including subgroup analyses, univariable and multivariable meta-regressions, and advanced sensitivity analyses such as GOSH/DBSCAN clustering. Furthermore, the use of SMDs allowed harmonization of IPFD metrics across different MRI protocols—including magnet strength, acquisition parameters, ROI placement, and post-processing algorithms—enhancing comparability of pooled estimates. Finally, we examined the influence of differences in BMI, LFC, and age on excess IPFD in T2DM using multivariable meta-regression, providing a more nuanced understanding of their contributions to the observed heterogeneity. However, the present meta-analysis has several limitations, many of which stem from the nature of the eligible studies. First, between-study heterogeneity remained considerable despite extensive exploratory analyses, likely reflecting differences in population characteristics and study design. Advanced structural heterogeneity analyses suggested that this heterogeneity was structural rather than driven by specific studies, and that the pooled estimates remained robust. Second, meta-regression analyses were conducted at the study-level, which may not accurately capture individual-level relationships, underscoring the need for future research using individual participant data. Third, the observational design of the included studies precludes causal inference regarding the relationship between IPFD and T2DM. Fourth, important clinical variables (e.g., glycemic control, duration of T2DM, and use of background medications) were inconsistently reported, limiting the ability to adjust for potential confounders and possibly contributing to residual heterogeneity. Fifth, most included studies were conducted within clinical or hospital-based settings, which may restrict the generalizability of the findings. Last, several included studies had small sample sizes per group, which may have contributed to variability in effect size estimates and between-study heterogeneity. From a clinical perspective, our results support excessive IPFD as a robust metabolic feature associated with T2DM. In light of the PANDORA hypothesis [62], this pattern suggests that IPFD assessment may contribute to refined cardiometabolic phenotyping and potentially inform future risk stratification strategies [4]. Importantly, the utility of routine IPFD assessment remains uncertain, and our findings should not be interpreted as supporting implementation in clinical practice at present. Such a paradigm shift would require evidence on cost-effectiveness and added value over existing cardiometabolic markers. In addition, the observed differences in IPFD between T2DM and non-diabetic controls highlight an understudied therapeutic dimension [71]. Interventions such as glucose-lowering therapies, bariatric surgery, and lifestyle modification have been suggested to reduce IPFD, though the evidence remains limited [19, 20, 24, 71, 72]. These observations provide a biologically plausible rationale bridging the link between IPFD reduction and improvements in lipotoxicity, β-cell function, and insulin responsiveness [24, 73]. While causality cannot be inferred, these considerations support the interpretation that mitigating IPFD may contribute, at least in part, to the reversibility of T2DM observed in some intervention studies [74, 75]. Several important questions for future investigation emerge from this work. Prospective longitudinal studies are needed to elucidate the temporality of IPFD during the natural history of T2DM, ideally using serial MRI assessments in conjunction with cardiometabolic outcomes. Incorporating IPFD as an imaging endpoint in clinical trials of metabolic diseases [76, 77] would also help clarify its relevance as a therapeutic target. In parallel, further research is needed to better characterize the biological pathways linking IPFD to systemic metabolic dysfunction. Standardization of MRI protocols for IPFD quantification also remains an important methodological priority [18, 26]. Understanding IPFD as part of multiorgan ectopic fat burden may improve prediction of T2DM complications and support more refined patient stratification in precision diabetes frameworks [78, 79]. Conclusion This systematic review, meta-analysis, and meta-regression provides a comprehensive synthesis of the association between T2DM and MRI-measured IPFD, quantifying the magnitude of IPFD differences and examining the contribution of BMI, LFC, and age. The findings suggest that higher IPFD may represent a distinct metabolic feature of T2DM, independent of obesity and hepatic steatosis. Recognizing the structural pancreatic changes associated with IPFD may help reframe T2DM not only as a disease of insulin resistance and β-cell dysfunction, but also as a condition characterized by an easily and objectively measurable morphological substrate in the pancreas—an insight with important implications for prevention, risk stratification, and therapeutic targeting in this population. Key References Petrov MS, Taylor R. Intra-pancreatic fat deposition: bringing hidden fat to the fore. Nat Rev Gastroenterol Hepatol. 2022;19(3):153–168. This landmark review provides a comprehensive discussion of the role of IPFD in health and disease. It also offers insights into the epidemiology, quantification, and lability of IPFD.Dong X, Zhu Q, Yuan C, et al. Associations of Intrapancreatic Fat Deposition With Incident Diseases of the Exocrine and Endocrine Pancreas: A UK Biobank Prospective Cohort Study. American Journal of Gastroenterology. 2024;119(6):1158–1166.Using UK Biobank data, this study shows longitudinal evidence on the association between high IPFD and the development of major exocrine and endocrine pancreatic diseases. It provides clinical evidence for positioning IPFD as a relevant therapeutic target to reduce the burden of pancreatic diseases.Chan TT, Tse YK, Lui RNS, et al. Fatty pancreas is independently associated with subsequent diabetes mellitus development: a 10-year prospective cohort study. Clin Gastroenterol Hepatol. 2022;20(9):2014–2022.e4.This prospective cohort study provides some of the first longitudinal evidence of an independent association between high IPFD and the development of diabetes mellitus.Zhang J, Liu Y, Petrov MS. Validation of magnetic resonance imaging for quantification of intrapancreatic fat deposition using phantom and histologic comparators: a systematic review and meta-analysis. Eur Radiol. Published online March 23, 2026. 10.1007/s00330-026-12475-x.This systematic review and meta-analysis demonstrates the high comparability of chemical shift-encoded MRI with both phantom models and histologic measurements for IPFD quantification.Agon HC, Shen Y, Petrov MS. Effects of glucose-lowering medications on intrapancreatic fat deposition: a systematic review and meta-analysis of randomized controlled trials. Obes Rev. 2026;27(6). 10.1111/obr.70087.This systematic review and meta-analysis demonstrates the reduction of IPFD following glucose-lowering medications. As a field-wide review, it calls for future research to investigate the effects of pipeline and approved drugs for metabolic diseases on IPFD. Petrov MS. Fatty change of the pancreas: the Pandora's box of pancreatology. Lancet Gastroenterol Hepatol. 2023;8(7):671–682.In this personal view, the PANDORA hypothesis is introduced as a comprehensive framework that positions IPFD as a central morphological and pathobiological determinant across major pancreatic diseases. Petrov MS, Taylor R. Intra-pancreatic fat deposition: bringing hidden fat to the fore. Nat Rev Gastroenterol Hepatol. 2022;19(3):153–168. This landmark review provides a comprehensive discussion of the role of IPFD in health and disease. It also offers insights into the epidemiology, quantification, and lability of IPFD. Petrov MS, Taylor R. Intra-pancreatic fat deposition: bringing hidden fat to the fore. Nat Rev Gastroenterol Hepatol. 2022;19(3):153–168. This landmark review provides a comprehensive discussion of the role of IPFD in health and disease. It also offers insights into the epidemiology, quantification, and lability of IPFD. This landmark review provides a comprehensive discussion of the role of IPFD in health and disease. It also offers insights into the epidemiology, quantification, and lability of IPFD. This landmark review provides a comprehensive discussion of the role of IPFD in health and disease. It also offers insights into the epidemiology, quantification, and lability of IPFD. Dong X, Zhu Q, Yuan C, et al. Associations of Intrapancreatic Fat Deposition With Incident Diseases of the Exocrine and Endocrine Pancreas: A UK Biobank Prospective Cohort Study. American Journal of Gastroenterology. 2024;119(6):1158–1166.Using UK Biobank data, this study shows longitudinal evidence on the association between high IPFD and the development of major exocrine and endocrine pancreatic diseases. It provides clinical evidence for positioning IPFD as a relevant therapeutic target to reduce the burden of pancreatic diseases. Dong X, Zhu Q, Yuan C, et al. Associations of Intrapancreatic Fat Deposition With Incident Diseases of the Exocrine and Endocrine Pancreas: A UK Biobank Prospective Cohort Study. American Journal of Gastroenterology. 2024;119(6):1158–1166.Using UK Biobank data, this study shows longitudinal evidence on the association between high IPFD and the development of major exocrine and endocrine pancreatic diseases. It provides clinical evidence for positioning IPFD as a relevant therapeutic target to reduce the burden of pancreatic diseases. Using UK Biobank data, this study shows longitudinal evidence on the association between high IPFD and the development of major exocrine and endocrine pancreatic diseases. It provides clinical evidence for positioning IPFD as a relevant therapeutic target to reduce the burden of pancreatic diseases. Using UK Biobank data, this study shows longitudinal evidence on the association between high IPFD and the development of major exocrine and endocrine pancreatic diseases. It provides clinical evidence for positioning IPFD as a relevant therapeutic target to reduce the burden of pancreatic diseases. Chan TT, Tse YK, Lui RNS, et al. Fatty pancreas is independently associated with subsequent diabetes mellitus development: a 10-year prospective cohort study. Clin Gastroenterol Hepatol. 2022;20(9):2014–2022.e4.This prospective cohort study provides some of the first longitudinal evidence of an independent association between high IPFD and the development of diabetes mellitus. Chan TT, Tse YK, Lui RNS, et al. Fatty pancreas is independently associated with subsequent diabetes mellitus development: a 10-year prospective cohort study. Clin Gastroenterol Hepatol. 2022;20(9):2014–2022.e4.This prospective cohort study provides some of the first longitudinal evidence of an independent association between high IPFD and the development of diabetes mellitus. This prospective cohort study provides some of the first longitudinal evidence of an independent association between high IPFD and the development of diabetes mellitus. This prospective cohort study provides some of the first longitudinal evidence of an independent association between high IPFD and the development of diabetes mellitus. Zhang J, Liu Y, Petrov MS. Validation of magnetic resonance imaging for quantification of intrapancreatic fat deposition using phantom and histologic comparators: a systematic review and meta-analysis. Eur Radiol. Published online March 23, 2026. 10.1007/s00330-026-12475-x.This systematic review and meta-analysis demonstrates the high comparability of chemical shift-encoded MRI with both phantom models and histologic measurements for IPFD quantification. Zhang J, Liu Y, Petrov MS. Validation of magnetic resonance imaging for quantification of intrapancreatic fat deposition using phantom and histologic comparators: a systematic review and meta-analysis. Eur Radiol. Published online March 23, 2026. 10.1007/s00330-026-12475-x.This systematic review and meta-analysis demonstrates the high comparability of chemical shift-encoded MRI with both phantom models and histologic measurements for IPFD quantification. This systematic review and meta-analysis demonstrates the high comparability of chemical shift-encoded MRI with both phantom models and histologic measurements for IPFD quantification. This systematic review and meta-analysis demonstrates the high comparability of chemical shift-encoded MRI with both phantom models and histologic measurements for IPFD quantification. Agon HC, Shen Y, Petrov MS. Effects of glucose-lowering medications on intrapancreatic fat deposition: a systematic review and meta-analysis of randomized controlled trials. Obes Rev. 2026;27(6). 10.1111/obr.70087.This systematic review and meta-analysis demonstrates the reduction of IPFD following glucose-lowering medications. As a field-wide review, it calls for future research to investigate the effects of pipeline and approved drugs for metabolic diseases on IPFD. Agon HC, Shen Y, Petrov MS. Effects of glucose-lowering medications on intrapancreatic fat deposition: a systematic review and meta-analysis of randomized controlled trials. Obes Rev. 2026;27(6). 10.1111/obr.70087.This systematic review and meta-analysis demonstrates the reduction of IPFD following glucose-lowering medications. As a field-wide review, it calls for future research to investigate the effects of pipeline and approved drugs for metabolic diseases on IPFD. This systematic review and meta-analysis demonstrates the reduction of IPFD following glucose-lowering medications. As a field-wide review, it calls for future research to investigate the effects of pipeline and approved drugs for metabolic diseases on IPFD. This systematic review and meta-analysis demonstrates the reduction of IPFD following glucose-lowering medications. As a field-wide review, it calls for future research to investigate the effects of pipeline and approved drugs for metabolic diseases on IPFD. Petrov MS. Fatty change of the pancreas: the Pandora's box of pancreatology. Lancet Gastroenterol Hepatol. 2023;8(7):671–682.In this personal view, the PANDORA hypothesis is introduced as a comprehensive framework that positions IPFD as a central morphological and pathobiological determinant across major pancreatic diseases. Petrov MS. Fatty change of the pancreas: the Pandora's box of pancreatology. Lancet Gastroenterol Hepatol. 2023;8(7):671–682.In this personal view, the PANDORA hypothesis is introduced as a comprehensive framework that positions IPFD as a central morphological and pathobiological determinant across major pancreatic diseases. In this personal view, the PANDORA hypothesis is introduced as a comprehensive framework that positions IPFD as a central morphological and pathobiological determinant across major pancreatic diseases. In this personal view, the PANDORA hypothesis is introduced as a comprehensive framework that positions IPFD as a central morphological and pathobiological determinant across major pancreatic diseases. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1