Sex differences in body composition in people with prediabetes and type 2 diabetes as compared with people with normal glucose metabolism: the Maastricht Study

Aims/hypothesis Obesity is a major risk factor for type 2 diabetes. However, body composition differs between women and men. In this study we investigate the association between diabetes status and body composition and whether this association is moderated by sex. Methods In a population-based cohort study (n=7639; age 40–75 years, 50% women, 25% type 2 diabetes), we estimated the sex-specific associations, and differences therein, of prediabetes (i.e. impaired fasting glucose and/or impaired glucose tolerance) and type 2 diabetes (reference: normal glucose metabolism [NGM]) with dual-energy x-ray absorptiometry (DEXA)- and MRI-derived measures of body composition and with hip circumference. Sex differences were analysed using adjusted regression models with interaction terms of sex-by-diabetes status. Results Compared with their NGM counterparts, both women and men with prediabetes and type 2 diabetes had more fat and lean mass and a greater hip circumference. The differences in subcutaneous adipose tissue, hip circumference and total and peripheral lean mass between type 2 diabetes and NGM were greater in women than men (women minus men [W–M] mean difference [95% CI]: 15.0 cm2 [1.5, 28.5], 3.2 cm [2.2, 4.1], 690 g [8, 1372] and 443 g [142, 744], respectively). The difference in visceral adipose tissue between type 2 diabetes and NGM was greater in men than women (W–M mean difference [95% CI]: −14.8 cm2 [−26.4, −3.1]). There was no sex difference in the percentage of liver fat between type 2 diabetes and NGM. The differences in measures of body composition between prediabetes and NGM were generally in the same direction, but were not significantly different between women and men. Conclusions/interpretation This study indicates that there are sex differences in body composition associated with type 2 diabetes. The pathophysiological significance of these sex-associated differences requires further study. Graphical abstract Supplementary Information The online version contains peer-reviewed but unedited supplementary material available at 10.1007/s00125-023-05880-0.

Because of a technical error (insufficient field of view on the first MR images) in 250 participants, the amount of SAT (n=250) and VAT (n=28) was inadvertently incomplete. In order to estimate the SAT and VAT in these participants as accurately as possible, a subset of participants with the highest waist circumference, yet without this issue was used to estimate (missing) SAT and VAT volume. In this subset of participants with complete MRI data, the amount of missing SAT and VAT was modeled based on age, sex and waist circumference, as well as the amount of VAT and SAT that was visible with incomplete settings. Herewith, actual SAT (ICC 0.978 (95%-CI 0.974; 0.982), n=433) and VAT (ICC 0.949 (0.901-0.974) n=35) could be estimated accurately.
Liver fat percentage was assessed through Dixon MR imaging (Supplemental Figure   1), as described in more detail elsewhere (6). In brief, after a scout scan, transversal twodimensional T2-weighted True Fast Imaging with steady-state-free precession (T2w TRUFI) images were acquired through the liver. Next, transversal two-dimensional turbo spin echo Dixon MR images were acquired through the liver during a breathhold. Three regions-ofinterest (ROIs) were drawn on the T2w TRUFI images of the liver by trained observers. Care was taken to position these ROIs avoiding visible structures in the liver, such as vessels and bile ducts, and in artifact-free regions. Subsequently, these ROIs were copied to the water and fat Dixon MR images to calculate the intrahepatic lipid fraction. This method was validated and calibrated against proton magnetic resonance spectroscopy ( 1 H-MRS), the gold standard to non-invasively quantify IHL, in 36 participants. With 1 H-MRS the CH2/H2O ratio was determined by calculating the ratio of the T2-corrected spectral peak areas of CH2 and water.

Assessment of glucose metabolism status
To determine glucose metabolism status (GMS), all participants underwent a standardized 2hour (2-h) 75 gram oral glucose tolerance test after fasting overnight. For safety reasons, participants using insulin or with a fasting glucose level above 11.0 mmol/L, as determined by a finger prick, did not undergo the oral glucose tolerance test. For these individuals, fasting glucose level and information about diabetes medication were used to determine GMS. GMS was defined according to the WHO 2006 criteria into normal glucose metabolism, impaired fasting glucose, impaired glucose tolerance (combined as prediabetes), and type 2 diabetes (7).
Participants on blood glucose lowering medication were classified as having type 2 diabetes.

Assessment of covariates and population characteristics
We used a questionnaire to assess age (years), sex, smoking status (never, current, former), alcohol use (g/day), adherence to the Dutch dietary guidelines and the indication of diet quality (based on fourteen out of fifteen components of the Dutch Healthy Diet index 2015, as information on filtered coffee intake was not collected (8)), calculated from a validated food frequency questionnaire (9) the total score ranges between 0 (no adherence) and 130 (complete adherence), educational level (low, intermediate, high), physical activity level (hours of moderate to vigorous physical activity per week) and postmenopausal status in women (10).
Women who indicated that they had had a menstrual period in the preceding 12 months were classified as premenopausal women; if they indicated that they had not had a menstrual period in the preceding 12 months they were classified as postmenopausal women. We assessed medication use, e.g. glucose-lowering, lipid-modifying and antihypertensive medication use, as well as postmenopausal hormone replacement therapy, during a medication interview where generic name, dose, and frequency were registered (10). Additionally, a variable was constructed which included medication known for weight gain or weight loss as possible side effect. It was defined as using one or more of the following drugs: hormonal contraceptives, antidepressants, antipsychotic drugs, lithium, medicinal cannabis, beta-blockers, anti-epileptics (i.e. valproic acid, gabapentin, carbamazepine and topiramate) mineralocorticoids (i.e. fludrocortisone), glucocorticoids (i.e. betamethasone, dexamethasone, methylprednisolone, prednisolone, prednisone, triamcinolone (acetonide), hydrocortisone or cortisone) (11). We determined HbA1c, fasting glucose, 2-h postload glucose, waist and hip circumference, BMI, triglyceride levels, total cholesterol, HDL cholesterol and systolic and diastolic office blood pressure as described elsewhere (10).

Statistical analyses
To investigate the robustness of the results obtained by the described analyses, we did several sensitivity analyses. In the imputed dataset (N=7639), a sensitivity analysis was conducted to investigate if time between baseline measurements and the DEXA and MRI scan (lag time), which were performed at a later moment in time, influenced the results. We first adjusted for lag time and additionally we repeated the original analyses in participants having a lag time ≤ 6 months (DEXA N=2129, MRI N=2834) and in participants having a lag time > 6 months (DEXA N=5510, MRI N=4805). Second, we repeated all analyses after exclusion of premenopausal women (N=809) and women in whom menopausal status was unclear (N=69; analysis population N=6761). Third, if body size prohibited the determination of the amount of SAT and VAT according to the study protocol, estimated values were used. We repeated the analyses after exclusion of participants with estimated values (N=250; analysis population N=4119) in the original (not imputed) dataset. We performed these analyses in the original dataset as the variables of SAT and VAT, including the estimated values, were used for imputation of the data. For these analyses, complete case analyses were performed. Finally, for all investigated associations, we performed complete case analysis in the original dataset to compare the results with the multiple imputation approach.  DEXA=dual-energy X-ray absorptiometry; MRI=magnetic resonance imaging.

Sex difference
Sex-specific differences are expressed as linear regression coefficients (95%-CI) of the dependent variables, which indicate mean differences (βs) or geometric mean ratios (GMRs) in amount of fat mass, lean mass, subcutaneous or visceral adipose tissue or liver fat percentage according to glucose metabolism status. The reference category for prediabetes and type 2 diabetes is normal glucose metabolism status. a GMRs Differences between sexes are expressed as linear regression coefficients (95%-CI) of the interaction terms sex*prediabetes and sex*type 2 diabetes, which indicate differences between women and men in mean differences (WM-βs) or women to men ratio of geometric mean ratios (WM-GMRs) in amount of fat mass, lean mass, subcutaneous or visceral adipose tissue or liver fat percentage according to glucose metabolism status. a WM-GMRs. Statistically significant differences between the sexes are typed in bold. *P value <0.05 Model 3: adjusted for age, height, physical activity, healthy diet score, educational level, alcohol consumption and smoking status and use of medication known for weight gain and/or loss as possible side effect † For each potential confounder included, an interaction term (sex by potential confounder) was incorporated in the same model † associations with liver fat percentage were not adjusted for height; associations with total and peripheral lean mass were additionally adjusted for total fat mass  DEXA=dual-energy X-ray absorptiometry; MRI=magnetic resonance imaging.

Sex difference
Differences between sexes are expressed as linear regression coefficients (95%-CI) of the interaction terms sex*prediabetes and sex*type 2 diabetes, which indicate differences between women and men in mean differences (WM-βs) or women to men ratio of geometric mean ratios (WM-GMRs) in amount of fat mass, lean mass, subcutaneous or visceral adipose tissue or liver fat percentage according to glucose metabolism status. †WM-GMRs. Statistically significant differences between the sexes are typed in bold. *P value <0.05 Model 3: adjusted for age, height, physical activity, healthy diet score, educational level, alcohol consumption and smoking status and use of medication known for weight gain and/or loss as possible side effect † For each potential confounder included, an interaction term (sex by potential confounder) was incorporated in the same model † associations with liver fat percentage were not adjusted for height; associations with total and peripheral lean mass were additionally adjusted for total fat mass

ESM Table 3 Differences within and between sexes in mean differences in measures of body composition according to glucose metabolism -exclusion of premenopausal women
Prediabetes β or GMR (95%-CI)