PurposeTo investigate the regional variability of adipose tissue triglyceride composition in vivo using 1H MRS, examining potential confounders and corrections for artifacts, to allow for adipose tissue spectrum estimation.Materials and Methods1H magnetic resonance (MR) stimulated echo acquisition mode (STEAM) spectra were acquired in vivo at 3T from 340 adult patients (mean age 48.9 years, range 21–79 years; 172 males, 168 females; mean body mass index [BMI] 34.0, range 22–49 kg/m2) with known or suspected nonalcoholic fatty liver disease (NAFLD) in deep (dSCAT), surface (sSCAT) subcutaneous adipose tissue, and visceral adipose tissue (VAT). Triglyceride composition was characterized by the number of double bonds (ndb) and number of methylene‐interrupted double bonds (nmidb). A subset of patients (dSCAT n = 80, sSCAT n = 55, VAT n = 194) had the acquisition repeated three times to examine the repeatability of ndb and nmidb estimation.ResultsMean ndb and nmidb showed significant (P < 0.0001) differences between depots except for dSCAT and sSCAT nmidb (dSCAT ndb 2.797, nmidb 0.745; sSCAT ndb 2.826, nmidb 0.737; VAT ndb 2.723, nmidb 0.687). All ndb and nmidb estimates were highly repeatable (VAT ndb ICC = 0.888, nmidb ICC = 0.853; sSCAT: ndb ICC = 0.974, nmidb ICC = 0.964; dSCAT: ndb ICC = 0.959, nmidb ICC = 0.948).ConclusionAdipose tissue composition can be estimated repeatably using 1H MRS and different fat depots have different triglyceride compositions.Level of Evidence: 2J. MAGN. RESON. IMAGING 2017;45:1455–1463