Factors associated with body weight gain and insulin-resistance: a longitudinal study

Obesity prevalence is rising globally and is commonly attributed to chronic positive energy balance. Whether insulin resistance and hyperinsulinemia contribute causally to obesity, beyond being consequences, remains unresolved. Prior studies in adults are inconclusive and pediatric longitudinal studies show mixed results. Insulin resistance underlies metabolic syndrome and related complications, making it clinically important to determine the temporal direction between insulin resistance and weight gain. This study aimed to longitudinally assess, in a general population cohort, the factors associated with 4-year body weight change, including serum insulin and HOMA-IR as a surrogate for insulin resistance.

The paper summarizes prior evidence indicating uncertainty about the causal role of insulin resistance in weight gain: adult cohort studies have not conclusively shown that hyperinsulinemia or insulin resistance predicts obesity. Pediatric studies report conflicting findings, with some indicating no predictive relationship while others suggest childhood insulin resistance predicts later adiposity gains. The clinical significance of insulin resistance in metabolic syndrome and its pharmacologic management (e.g., biguanides) is noted, underscoring the need to clarify directionality between insulin resistance and obesity.

Design and cohort: Longitudinal observational single-center study within the ABCD (Nutrition, Cardiovascular Wellness, and Diabetes) Project (ISRCTN15840340), recruiting a cohort representative of Palermo, Italy. The ABCD_1 cohort was recruited in 2011 (ages ≥18 years). Participants were recontacted in 2015 and re-examined between March 21 and July 31 at the University of Palermo. The ABCD_2 cohort demographics did not differ significantly from ABCD_1. Individuals with known diabetes were excluded. Ethics approval was obtained (Palermo 1, Policlinico "P. Giaccone" University Hospital, 11/03/2014, ref: 3/2015); written informed consent was provided. Dietary assessment: Habitual energy and macronutrient intake over the preceding 12 months was assessed using a validated medium-length Food Frequency Questionnaire for the local Sicilian population. Adherence to the Mediterranean diet was quantified by the MEDI-LITE score. Dietary glycemic index and glycemic load were computed. Physical activity: Assessed by a locally validated questionnaire with 4 activity levels (1–4). Anthropometry and body composition: Height, weight (light clothing, no shoes), and BMI (kg/m²) were measured. Waist (umbilicus) and hip (most prominent buttock) circumferences were recorded. Blood pressure (two seated measures 5 min apart) and heart rate were measured (Omron M6). Body composition (fat mass, fat-free mass) was estimated by bioelectrical impedance analysis (Akern BIA-101 Anniversary) following manufacturer’s equations. Laboratory assays: Fasting plasma glucose, total cholesterol, HDL-C, triglycerides, uric acid, creatinine were measured with standard clinical chemistry methods. Basal insulin (Elecsys), hs-CRP, and HbA1c were measured. LDL-C was calculated by Friedewald’s formula; eGFR by CKD-EPI. HOMA-IR was calculated from fasting glucose and insulin. DNA was purified and genotyped for PNPLA3 rs738409 and TM6SF2 rs58542926 using TaqMan assays. Definitions: Normal-weight BMI <25 kg/m²; overweight/obese BMI ≥25 kg/m². Type 2 diabetes and prediabetes defined per consensus statements (2014 ADA). Weight change categories: weight gainers (G) defined as body weight increase ≥5 kg from 2011 to 2015; non-gainers (NG) otherwise. Sample flow: From ABCD_1 n=1461, exclusions included systemic/organ diseases, type 1 or 2 diabetes, loss to follow-up/refusal; ABCD_2 n=738, with further exclusions for incomplete labs and T2D, yielding n=707 for analysis (G n=87; NG n=620). Statistics: Continuous variables reported as mean±SD; categorical as percentages. Between-group comparisons used unpaired Student’s t-tests; within-individual changes used paired t-tests; categorical differences by chi-squared. Pearson correlations assessed associations between continuous variables. For regression, absolute changes (Δ) were dichotomized (0 if Δ≤0; 1 if Δ>0). Multiple logistic regression predicted odds of being a gainer (0/1). Multiple stepwise linear regression modeled final HOMA-IR. Significance threshold P<0.05. Analyses were performed with Systat v13.0 (Windows).

• Cohort: 707 adults without diabetes at follow-up; 87 gainers (G, ≥5 kg gain over 4 years; ~12%) and 620 non-gainers (NG).
• Baseline comparability: G and NG had similar initial BMI (27.8±6.5 vs 28.1±5.1 kg/m²), body weight (76.7±22.1 vs 74.2±14.7 kg), insulin, HOMA-IR, and energy intake; G were younger (44±13 vs 51±13 years, P<0.001).
• Weight change: Over 4 years, G gained ~9.6 kg on average (76.7±22.1 to 86.3±23.7 kg, P<0.001; BMI 27.8±6.5 to 31.5±7.2, P<0.001), while NG lost modest weight (74.2±14.7 to 72.9±14.2 kg, P<0.001; BMI 28.1±5.1 to 27.6±5.0, P<0.001). Waist circumference increased in G (95±17 to 103±15 cm, P<0.001) and was unchanged in NG.
• Diet: Baseline EI similar. At follow-up, G increased EI (1383±337 to 1444±392 kcal/day, P<0.001) while NG decreased EI (1434±532 to 1325±353 kcal/day, P<0.001). Both groups reduced fat intake and increased carbohydrate intake; protein intake stable. MEDI-LITE score improved in NG (9.9±1.9 to 10.3±2.1, P<0.001) but not in G.
• Metabolic markers: In G, fasting insulin rose (9.3±5.1 to 12.9±7.2 µU/mL, P<0.001) and HOMA-IR increased (2.07±1.21 to 2.94±1.75, P<0.001). NG showed no significant change in insulin or HOMA-IR. Fasting glucose increased slightly in G (87±10 to 90±11 mg/dL, P<0.05). Lipid changes were generally favorable in NG.
• Genetics: PNPLA3 rs738409 CG/GG alleles were more prevalent in G than NG (CG+GG 56.1% vs 43.0%; GG 15.7% vs 6.6%; P<0.05). TM6SF2 rs58542926 allele distribution did not differ.
• Correlations with weight change (Δ): Δ weight correlated with Δ insulin (r=0.36, P<0.001), Δ HOMA-IR (r=0.32, P<0.001), Δ MEDI-LITE score (r=-0.08, P<0.05), Δ EI (r=0.17, P<0.001), Δ dietary glycemic index (r=0.08, P<0.05), Δ glycemic load (r=0.14, P<0.001), and age (r=-0.19, P<0.001).
• Predictors of being a gainer (logistic regression): Age OR 1.031 (95% CI 1.022–1.040, P<0.001); increase in EI (Δ EI>0 vs ≤0) OR 2.257 (1.345–3.789, P<0.005); PNPLA3 CG/GG alleles OR 1.700 (1.066–2.709, P<0.05). Δ glycemic load and Δ MEDI-LITE were not significant in the multivariable model.
• Determinants of final insulin resistance: Final HOMA-IR correlated with age, sex, BMI, waist circumference, waist-to-hip ratio, fat-free mass, fat mass percentage, and EI. In multivariable stepwise regression, only final BMI (P<0.001), waist circumference (P<0.001), and EI (P<0.05) remained independently associated: HOMA-IR = 3.13 + (0.067 × BMI) + (0.033 × waist circumference) + (0.0001 × energy intake); R²=0.31 (P<0.001).

Energy intake emerged as the main modifiable determinant of weight gain over 4 years. Despite similar baseline anthropometry and diet, gainers increased EI and gained substantial weight, while non-gainers modestly reduced EI and weight. Lifestyle advice provided in the initial ABCD study may have aided improvements in diet quality among non-gainers (higher MEDI-LITE score, lower glycemic index), underscoring the potential of individualized prevention strategies. Genetically, the PNPLA3 rs738409 CG/GG variants were independently associated with becoming a gainer, suggesting a possible biological mechanism linking impaired lipid droplet autophagy (lipophagy) with susceptibility to positive energy balance and fat accumulation. This aligns with literature on PNPLA3’s role in hepatic lipid handling but requires further study for causality. Regarding insulin resistance, the temporal data indicate IR rose after weight gain rather than preceding it. Final HOMA-IR was independently linked to adiposity (BMI, waist) and EI, supporting the concept that IR develops secondary to excess energy substrate and may function as a protective cellular response against osmotic and oxidative stress from nutrient overload. While IR may mitigate intracellular damage, persistent IR contributes to metabolic complications, suggesting that reducing EI (and using anti-obesity pharmacotherapy when needed) is a rational approach rather than attempting to pharmacologically abolish IR per se.

Excessive energy intake is primarily associated with clinically meaningful body weight gain over 4 years. Genetic predisposition, particularly PNPLA3 rs738409 CG/GG variants, may favor a positive energy balance and weight gain. Insulin resistance increases following weight gain and correlates with energy intake and central adiposity, consistent with a protective response to intracellular nutrient overload. Strategies to prevent and treat obesity and insulin resistance should prioritize reducing energy intake and improving diet quality, with consideration of adjunctive anti-obesity medications when dietary measures alone are insufficient. Future research should expand genetic analyses and employ gold-standard measures of insulin sensitivity to better elucidate causality and mechanisms.

• Genetic scope was limited to PNPLA3 and TM6SF2; other loci related to β-cell function and insulin resistance were not assessed.
• Insulin resistance was estimated by HOMA-IR rather than euglycemic hyperinsulinemic clamp, the gold standard; although correlated, clamp studies provide more precise measures.
• Body composition by BIA was not performed at baseline (ABCD_1), limiting analysis of composition changes over time; only follow-up values were available.
• As an observational study, residual confounding and selection bias (loss to follow-up) are possible despite efforts to maintain cohort representativeness.