Background Substantial weight reduction is often associated with loss of muscle mass. Tirzepatide has been associated with significant reductions in body weight in type 2 diabetes trials and a beneficial effect on body fat distribution in the SURPASS-3 MRI substudy. This post-hoc exploratory analysis studied the association of tirzepatide treatment with changes in thigh muscle volume, muscle volume Z score, and muscle fat infiltration, and aimed to contextualise the results using longitudinal MRI data from UK Biobank participants. Methods SURPASS-3 was a randomised, open-label, parallel-group, phase 3 trial. The multicentre (45 sites) and multinational (eight countries) MRI substudy of SURPASS-3 enrolled insulin-naive adults (aged >= 18 years) with type 2 diabetes who were on treatment with metformin with or without a sodium-glucose co-transporter-2 (SGLT-2) inhibitor, had an HbA1c of 7<middle dot>0-10<middle dot>5% (53-91 mmol/mol), a BMI of at least 25 kg/m(2), and a fatty liver index of at least 60. Participants were randomly assigned (1:1:1:1) to receive subcutaneous injection once per week of tirzepatide (5, 10, or 15 mg), or subcutaneous injection once per day of titrated insulin degludec. Thigh muscle fat infiltration, muscle volume, and muscle volume Z score (invariant to sex, height, weight, and BMI) were quantified by MRI at baseline and week 52. In this post-hoc analysis, we assessed the differences between mean baseline and week 52 muscle composition values in the tirzepatide groups (pooled 5 mg, 10 mg, and 15 mg group, and per dose group) and insulin degludec group using paired t tests, and the differences in muscle composition changes with pooled tirzepatide versus insulin degludec via adjusted ANCOVA models. Observed changes in muscle fat infiltration, muscle volume, and muscle volume Z scores were compared using paired t tests to population-based estimates calculated from multiple linear regression models fitted to UK Biobank data (n=2942), capturing associations with change in body weight. Analyses were done by modified intention to treat, in the participants enrolled in the MRI substudy with a valid MRI scan at week 52. The SURPASS-3 clinical trial is registered with ClinicalTrials.gov, NCT03882970, and is complete. Findings Participants were assessed for eligibility and recruited from April 1, 2019, to Nov 15, 2019. Among 502 participants assessed for eligibility to participate in the MRI substudy, 296 were enrolled, and 246 had a valid week 52 MRI scan and were included in the post-hoc analyses (tirzepatide 5 mg, n=63; tirzepatide 10 mg, n=60; tirzepatide 15 mg, n=67; insulin degludec, n=56; 147 [59<middle dot>8%] male participants and 99 [40<middle dot>2%] female participants). At baseline, overall mean age was 56<middle dot>0 years (SD 9<middle dot>9), median duration of type 2 diabetes was 6<middle dot>7 years (IQR 3<middle dot>7 to 10<middle dot>7), mean HbA1c was 8<middle dot>3% (SD 0<middle dot>9), mean BMI was 33<middle dot>4 kg/m2 (SD 4<middle dot>8), and 76 (30<middle dot>9%) were on an SGLT-2 inhibitor. Mean baseline muscle fat infiltration, muscle volume, and muscle volume Z scores were similar between the pooled tirzepatide group and insulin degludec group. For the pooled and individual tirzepatide dose groups, significant reductions were observed from baseline to week 52 in muscle fat infiltration (for pooled tirzepatide, mean change-0<middle dot>36 percentage points [95% CI-0<middle dot>48 to-0<middle dot>25], p<0<middle dot>0001), muscle volume (-0<middle dot>64 L [95% CI-0<middle dot>74 to-0<middle dot>54], p<0<middle dot>0001), and muscle volume Z score (-0<middle dot>22 [95% CI-0<middle dot>29 to-0<middle dot>15], p<0<middle dot>0001), which occurred in the context of significant weight reduction. Insulin degludec was associated with a modest and significant increase in bodyweight and muscle volume, but no significant change in the other variables. The changes in all three muscle composition variables with pooled tirzepatide were significantly different compared to those with insulin degludec. In tirzepatide-treated participants, observed muscle volume changes across all tirzepatide doses were similar to population-based estimated changes (for pooled tirzepatide, mean difference vs population-based estimate, -0<middle dot>04 L [95% CI-0<middle dot>11 to 0<middle dot>03], p=0<middle dot>22); whereas, observed reductions in muscle fat infiltration across all doses were significantly greater than population-based estimates (for pooled tirzepatide, mean difference-0<middle dot>42 percentage points [95% CI-0<middle dot>54 to-0<middle dot>31], p<0<middle dot>0001), and the observed reduction in muscle volume Z score with tirzepatide 15 mg was significantly greater than the population-based estimate (mean difference-0<middle dot>18 [95% CI-0<middle dot>29 to-0<middle dot>07], p=0<middle dot>0016). Interpretation In the SURPASS-3 MRI substudy, in the context of significant improvements in bodyweight and fat distribution, tirzepatide treatment was associated with potentially favourable changes in muscle fat infiltration and reductions in muscle volume broadly in accordance with the general association between changes in muscle volume and bodyweight. The present findings provide additional information on the potential effect of tirzepatide on muscle health that might help health-care providers when deciding among treatment options for individual patients. Copyright (c) 2025 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY-NC-ND 4.0 license.