Impact of baseline adipose tissue characteristics on change in adipose tissue volume during a low calorie diet in people with obesity—results from the LION study

Obesity-related comorbidities such as diabetes, cardiovascular disease, and certain cancers can be mitigated by weight-loss–focused lifestyle interventions, ideally reducing excess adipose tissue (AT) and ectopic fat. Different fat depots (subcutaneous, visceral) have distinct metabolic profiles and functional differences, making depot-specific changes during weight loss clinically relevant. Short-term weight loss predicts long-term maintenance, yet individual variability is substantial and influenced by adherence, insulin resistance, genetics/epigenetics, gut microbiota, sleep, and basal metabolic rate; age and sex effects are inconsistent. AT distribution affects intervention success: abdominal obesity with higher visceral adipose tissue (VAT) often benefits more than gluteal-femoral obesity dominated by subcutaneous adipose tissue (SAT). Prior work suggests baseline VAT, VAT/SAT ratio, or higher total AT may associate with greater weight and VAT loss, but many studies emphasized changes rather than predictive baselines, had small samples, targeted single depots, or relied on single-slice imaging rather than volumetric approaches. MRI offers accurate, radiation-free, volumetric AT phenotyping with proton density fat fraction (PDFF) quantification. This study aimed to assess changes in AT and apparent lipid volumes during an 8-week formula-based LCD and to evaluate baseline parameters for their correlation with and prognostic value for short-term AT loss in adults with obesity.

The literature indicates depot-specific metabolic profiles of AT and links between AT distribution and weight-loss outcomes. Imaging-based studies have associated greater baseline VAT, VAT/SAT ratio, or total AT with larger weight and VAT reductions, yet many prior works focused on change rather than predictive baselines, used small cohorts, single depots, or single-slice methods that poorly predict volumetric changes. Volumetric MRI or CT provides more accurate depot quantification, and Dixon-based chemical shift encoding MRI with PDFF is considered the gold standard for spatially resolved fat quantification. Prior interventions frequently report greater absolute SAT loss than VAT, though relative changes can vary; sex dimorphism in fat distribution and loss patterns is established. This study builds on these findings by using 3D MRI to quantify whole abdominopelvic AT volumes, regionalizing them into thirds, and integrating PDFF-derived apparent lipid volumes to examine both change and predictive baseline characteristics.

Study design and participants: From the LION lifestyle intervention trial, 127 adults with obesity (BMI 30.0–39.9 kg/m²) were recruited between Oct 2019 and Oct 2021 for abdominal–pelvic MRI; 81 (49 females; mean age 46.3 years) completed baseline and post-intervention scans following an 8-week formula-based low-calorie diet (LCD) of 800 kcal/day plus optional 200 g non-starchy vegetables. Ethical approval was obtained (TUM School of Medicine and Health; Project 69/195; NCT04023942), and written informed consent was provided. Inclusion/exclusion criteria followed the LION protocol, with standard MRI contraindications added. Anthropometry: Trained staff measured waist and hip circumferences; WHR was calculated. Weight (fasted, light clothing, −1 kg adjustment) and body fat% were obtained by bioimpedance (Tanita BC-418MA). Height was measured by stadiometer, and BMI computed. Baseline and follow-up anthropometry were scheduled near MRI (baseline interval mean 8 days; follow-up mean 1.7 days). MRI acquisition: 3T scanner (Ingenia Elition X, Philips, software 5.6) with 16-channel torso and table posterior coils. A 6-echo multi-echo gradient echo sequence with bipolar gradients was acquired in four breath-hold stacks (10.3 s each), covering liver dome to femoral head center. PDFF mapping: Vendor complex-based fat quantification (Philips mDIXON Quant) accounting for multi-peak fat spectrum, single T2* correction, and phase errors produced PDFF maps. Segmentation and regionalization: Automated deep learning-based segmentation of SAT and VAT from liver dome to middle femoral heads was performed using an established pipeline (nnU-Net/FatSegNet-based). SAT and VAT volumes and mean PDFF were extracted. To assess craniocaudal distribution, each compartment was subdivided into equidistant upper, middle (periumbilical), and lower (pelvic/gluteal to hip joint) thirds. Apparent lipid volume was calculated as PDFF × volume for SAT and VAT (acknowledging MR-invisible components are not included). Absolute losses were defined as follow-up minus baseline volumes (L); relative losses as (follow-up − baseline)/baseline × 100%. VAT/SAT ratio = VAT volume/SAT volume. Volumes were normalized by the segmented region length (L/cm) to account for physique differences. Statistical analysis: Normality was checked with cumulative frequency plots. Paired t-tests compared baseline vs follow-up; independent t-tests assessed sex and depot differences. Standardized mean differences (Cohen’s d) quantified effect sizes for normalized subvolumes (overall and by sex). Correlations used Pearson coefficients. Stepwise multiple linear regression identified baseline predictors of relative total volume loss (Δ ATTV%) and relative apparent lipid volume loss (Δ ATLV%) for SAT and VAT, entering the five strongest correlates plus age and sex as covariates. Two-tailed α = 0.05; no multiple-testing correction due to exploratory design. Software: MedCalc v20.118.

- Cohort and weight loss: 81 participants (49F), mean baseline BMI 34.08 ± 2.75 kg/m², mean age 46.3 ± 10.97 years. After 8-week LCD, mean weight change −11.61 ± 3.07 kg (p < 0.01); BMI and body fat% decreased significantly. - Absolute vs relative depot changes: Absolute SAT total volume loss (ΔSATTV) −3.24 ± 1.07 L and SAT apparent lipid volume loss (ΔSATLV) −3.14 ± 1.03 L; absolute VAT total volume loss (ΔVATTV) −1.24 ± 0.66 L and VAT apparent lipid volume loss (ΔVATLV) −1.17 ± 0.64 L. Absolute SAT loss exceeded VAT loss (p < 0.01). Relative losses: SATTV −21.46 ± 6.73% (SATLV −23.10 ± 7.44%), VATTV −21.79 ± 6.99% (VATLV −25.83 ± 8.09%). Relative apparent lipid loss was higher in VAT than SAT (p < 0.01). - PDFF changes: PDFF decreased in both SAT and VAT, with a stronger decrease in VAT (p < 0.01). - Regional effects (thirds): Significant reductions in all subvolumes. Largest effects were in the lower third for both SAT and VAT. Standardized mean differences (SMD): overall SAT lower third d = −0.81 (large), VAT lower third d = −0.71 (medium); females: SAT lower third d = −0.89 (large); males: SAT lower third d = −0.81 (large), VAT middle d = −0.85 and lower d = −0.86 (both large). - Sex differences: Men had larger absolute losses than women (e.g., males: ΔSATTV −3.50 ± 1.23 L; ΔVATTV −1.76 ± 0.68 L; females: ΔSATTV −3.07 ± 0.93 L; ΔVATTV −0.91 ± 0.37 L). Relative losses were higher in men across compartments. - Correlations: Relative SATTV loss correlated most with normalized baseline SAT volume in the lower third (r = 0.52, p < 0.01). Other strong correlates included normalized total SAT volume (r = 0.47), normalized SAT apparent lipid volume (r = 0.47), and body fat% (r = 0.49) (all p < 0.01). Relative VATTV loss correlated best with normalized SAT volume in the lower third and total normalized SAT volume (both r = 0.48, p < 0.01). Relative VATLV loss correlated with normalized total SAT volume and normalized SAT apparent lipid volume (r = 0.53 each), and SAT lower and upper thirds (r = 0.52 and r = 0.50) (all p < 0.01). Baseline SAT PDFF correlated with all loss measures; baseline VAT PDFF did not. - Regression predictors: For ΔSATTV%: normalized baseline lower-third SAT volume (b = 0.24, p < 0.01) and body fat% (b = 0.003, p = 0.02). For ΔSATLV%: normalized baseline lower-third SAT volume (b = 0.28, p < 0.01) and body fat% (b = 0.003, p = 0.01). For ΔVATTV%: normalized baseline lower-third SAT volume (b = 0.32, p < 0.01). For ΔVATLV%: normalized total SAT volume (b = 1.03, p < 0.01) and normalized middle-third SAT volume (b = −0.49, p = 0.04). Smaller lower-third SAT volumes favored greater SAT and VAT losses. - Weight/BMI loss correlations: Overall, relative weight and BMI losses did not correlate with baseline MRI/anthropometry. In females only, higher relative weight/BMI loss correlated with lower baseline body fat% (r ≈ 0.37–0.38), smaller waist circumference (r = 0.30, p = 0.04), and smaller lower-third SAT volumes (r = 0.32, p ≈ 0.02–0.03).

This study demonstrates significant reductions in anthropometric and MRI-derived AT measures after an 8-week LCD in adults with obesity, addressing the research question of how baseline AT features relate to short-term AT loss. Absolute losses were larger in SAT than VAT, while relative apparent lipid losses were higher in VAT, suggesting depot-specific metabolic responses: early preferential mobilization of VAT lipids under caloric restriction may reflect more responsive lipid metabolism and sympathetic drive in VAT. Regional analysis revealed the largest changes in the lower abdominopelvic third for both SAT and VAT, a finding enabled by volumetric MRI and regional subdivision that differs from traditional anthropometric perspectives (standing vs supine distribution). Crucially, normalized baseline SAT volume in the lower third emerged as the most consistent predictor of both SAT and VAT losses, with smaller volumes predicting greater reductions. This implies that lower gluteal-pelvic SAT burden may facilitate VAT mobilization and overall AT reduction during short-term LCD, aligning with prior observations that higher VAT or higher VAT/total AT ratio predicts successful weight loss. Baseline SAT PDFF’s inverse association with loss suggests tissue composition influences responsiveness, potentially reflecting adipocyte lipid content or hydration states. The results underscore sex dimorphism: men experienced greater absolute and, often, relative depot losses, particularly in VAT middle and lower regions, consistent with android vs gynoid distributions. Collectively, the findings highlight the value of 3D MRI phenotyping, including regionalized volumetrics and PDFF-based apparent lipid volumes, to predict and monitor intervention response, potentially informing individualized strategies focused on depot-specific risk and expected responsiveness.

An 8-week low-calorie diet in adults with obesity led to significant, sex- and depot-specific reductions in adipose tissue, with the largest regional losses in the lower abdominopelvic third for both SAT and VAT. The normalized SAT volume in the lower third at baseline was the strongest predictor of subsequent SAT and VAT losses, with smaller baseline volumes favoring greater reductions. Measuring lower-third SAT volume by MRI could help predict short-term AT-loss success during LCD. Future work should evaluate long-term outcomes (e.g., 12 months and beyond), weight cycling, and validate predictive markers across broader BMI ranges and diverse populations.

- PDFF partial volume effects: PDFF cannot distinguish intracellular water from non-lipid tissues within a voxel, potentially affecting accuracy. - Apparent lipid volume estimation: Calculation (PDFF × volume) excludes MR-invisible components (e.g., water bound to macromolecules), hence represents “apparent” lipid content. - Timing differences: There was a time gap between anthropometric and MRI measurements (baseline mean 8 days; follow-up mean 1.7 days). - Anatomical coverage: Segmentations extended only to the middle of the femoral head; comparisons with studies including full gluteofemoral/leg depots may be limited, and supine MRI redistributes AT cranially vs standing anthropometry. - No control group: All participants underwent the same 8-week LCD; causal attribution beyond the intervention period is limited. - Normalization choice: Volumes were normalized by segmented region length; alternative normalizations (BMI, height, body surface area) might yield different inferences, though length provided the clearest results here. - Regression not sex-stratified: Multiple regression models were not run separately by sex due to sample size relative to covariates.