Adiposity and diabetes in Mexican adults: evidence from observational and mendelian randomization analyses of a large prospective study

Background Obesity is a major risk factor for diabetes and vascular disease, but different fat distribution patterns may reflect distinct biological mechanisms. In people with established diabetes and persistent hyperglycaemia, reverse causation from diabetes-related weight loss may distort observational associations. We assessed the causal effects of multiple adiposity traits on diabetes phenotypes and vascular-metabolic mortality in Mexican adults, comparing observational and Mendelian randomization (MR) estimates. Methods We analysed 126,561 participants aged 35-74 years from the Mexico City Prospective Study. Measured and genetically predicted adiposity traits included body mass index (BMI), predicted fat mass percentage (FM%), waist and hip circumference. Diagnosed diabetes was defined by self-report or use of anti-diabetic treatment. Among diagnosed cases, HbA1c ≥9% indicated poor glycaemic control. Undiagnosed diabetes was defined as HbA1c ≥6.5% without prior diagnosis. Vascular-metabolic mortality included vascular, renal, hepatobiliary, and acute diabetic crisis deaths. Causal effects were estimated using one sample MR with multi-ancestry genetic instruments; multivariable MR assessed independent effects of each marker. Results Participants had a mean age of 51 years and 68% were women; 13% had diagnosed diabetes and 5% undiagnosed diabetes, with half of diagnosed cases poorly controlled. During follow-up, 7086 vascular-metabolic deaths occurred at ages 35-74 years. Observationally, undiagnosed diabetes increased progressively with all adiposity markers. In contrast, associations with diagnosed diabetes were attenuated and flat or U-shaped for BMI and hip circumference, and inverse for poor glycaemic control, consistent with reverse causation. Genetic instruments explained 2-3% of adiposity variation, with similar strength by sex. Genetically predicted adiposity was strongly associated with diagnosed diabetes: odds ratios per SD were 2.49 (95% CI 2.24–2.75) for BMI, 2.30 (2.04–2.61) for FM%, 3.04 (2.69–3.44) for waist circumference, and 1.63 (1.48–1.80) for hip circumference. Associations were stronger for poorly controlled diabetes. In multivariable MR, estimates were attenuated by one third for BMI and waist circumference, by nearly two thirds for FM%, and were null for hip circumference. Associations for undiagnosed diabetes were similar but weaker. Waist circumference showed the strongest association with vascular-metabolic mortality (HR per SD increase 2.38, 95% CI 2.02-2.81), largely independent of other markers (adjusted HR 2.19, 1.82-2.64). The association was halved by adjustment for genetic diabetes predisposition. Conclusion Observational analyses relating adiposity to diabetes in this population appear biased by reverse causation. MR analyses indicate that adiposity, particularly abdominal adiposity, is a major causal determinant of diabetes and vascular-metabolic mortality in Mexican adults.