Atherogenic index of plasma and its 5-year changes associated with type 2 diabetes risk: a 10-Year cohort study | Cardiovascular Diabetology

Baseline characteristics of the study participants

The demographic and clinical characteristics of participants are shown in Table 1. Among 14,968 participants included in this prospective cohort study, 2,165 developed T2DM during a median follow-up of 10.4 years. At baseline, the median age was 51, with 39.93% males. Participants who developed T2DM showed a higher proportion of those who were females (63.23%), those married (92.24%), those with education levels below high school (91.50%), individuals with a family history of diabetes (9.98%), and those who had used antihypertensive medications (15.47%) compared to those without T2DM (all p < 0.05). Those who developed T2DM showed higher levels of WC, BMI, SBP, DBP, FPG, TC, TG, LDL-C, and AIP compared to those without T2DM (all p < 0.001).

Table S2 shows baseline characteristics by quartiles of the AIP. A gradual increase in the proportion of males (from 36.93% in Q1 to 41.58% in Q4), smokers (from 25.68% to 30.19%), and alcohol drinkers (from 10.53% to 13.58%) is evident. The use of antihypertensive drugs also showed an increasing trend, rising from 7.30% in Q1 to 14.99% in Q4. Waist circumference increased from 75.90 cm in Q1 to 87.00 cm in Q4, while BMI rose from 22.20 to 25.48 kg/m². Similarly, systolic and diastolic blood pressures, fasting plasma glucose, total cholesterol, and triglycerides showed progressive increases across AIP quartiles. Conversely, HDL-C levels declined from 1.36 in Q1 to 0.97 mmol/L in Q4 (all p < 0.001).

Association between baseline AIP and 10-year T2DM risk

Table 2; Fig. 1(A) present the results of Cox proportional hazards models and RCS analysis for the association and dose-response relationship between AIP and the risk of T2DM. In the unadjusted model (Model a), compared to Q1 of AIP, the HRs of T2DM risk for Q2, Q3, and Q4 were 1.49 (95% CIs: 1.29–1.73), 2.04 (95% CIs: 1.77–2.34), and 3.36 (95% CIs: 2.95–3.83), respectively. After adjusting for gender, age, education, marital, income, smoking status, alcohol consumption, physical activity, family history of diabetes, BMI, systolic blood pressure and TC, the HRs of T2DM risk for Q2, Q3, and Q4 remained significant at 1.17 (95% CIs: 1.00–1.38), 1.38 (95% CIs: 1.18–1.62), and 1.96 (95% CIs: 1.68–2.29), respectively, with a significant linear trend across quartile groups (p for trend < 0.0001). Additionally, each 1-standard deviation (SD) increase in AIP was associated with a 33% increased risk of T2DM (aHR: 1.33; 95% CI: 1.26–1.39). The RCS analysis visually corroborates these findings, demonstrating a linear dose-dependent increase of baseline AIP with T2DM risk (p for nonlinearity = 0.927).

Using receiver-operating characteristic analysis with the Youden-index method, we identified an optimal AIP cut-off of 0.156. Participants were therefore classified into a high-AIP group (AIP ≥ 0.156) and a low-AIP group (AIP < 0.156). In multivariable models, individuals in the high-AIP group exhibited a 63% higher risk of incident T2DM than those in the low-AIP group (adjusted HR = 1.632, 95% CI 1.48–1.80). To assess the clinical utility of adding AIP to conventional risk prediction models, we conducted NRI and IDI analyses (Table S3). Even in the fully adjusted clinical model containing comprehensive traditional risk factors, AIP provided clinically meaningful incremental value with NRI of 0.237 (p < 0.001) and IDI of 0.0113 (p < 0.001).

Figure 2(A) illustrates the HRs and 95% CIs for the risk of T2DM among subgroups stratified by gender, age, smoking status, alcohol consumption, physical activity, and family history of diabetes. The positive association between baseline AIP and T2DM risk was consistently observed across different age groups (< 60 years: aHRs 1.33, 95% CI 1.26–1.40; ≥60 years: aHRs 1.34, 95% CI 1.20–1.49), genders (male: aHRs 1.28, 95% CI 1.18–1.39; female: aHRs 1.34, 95% CI 1.26–1.42), and smoking status (non-smoker: aHRs 1.34, 95% CI 1.27–1.42; smoker: aHRs 1.27, 95% CI 1.15–1.40). Similar patterns were observed for alcohol consumption, physical activity, and family history of diabetes. No significant multiplicative interactions were found between baseline AIP and any of the demographic or lifestyle factors (all p for interaction > 0.05), indicating that the association between baseline AIP and T2DM risk was homogeneous across these subgroups. Following this initial subgroup analysis, we conducted a stratified analysis based on baseline AIP levels (Table S4) which showed that within each subgroup the HRs for T2DM consistently increased from Q1 to Q4, with the highest risk observed in Q4 (p for trend < 0.0001).

Subgroup Analysis of the Association Between Baseline AIP (A) and 5-year Changes in AIP (B) with

To test the robustness of our findings regarding baseline AIP, sensitivity analyses were conducted using four additional models (Models e–h) (Table S5). Excluding participants with cardiovascular disease, cancer, and renal failure in Model e, the T2DM risk increased by 48% per 1-SD increase in AIP (aHR: 1.48; 95% CI: 1.41–1.55). After excluding those with impaired fasting glucose in Model f, the risk was 44% (aHR: 1.44; 95% CI: 1.36–1.52). Excluding participants with dyslipidemia in Model g, the risk increase was 35% (aHR: 1.35; 95% CI: 1.26–1.43). Finally, in Model h, excluding individuals who developed T2DM within 1 year, the risk increased by 47% (aHR: 1.47; 95% CI: 1.40–1.54).

We performed mediation analysis to examine whether WC mediates the association between AIP and T2DM risk (Figure S3 and Table S6). In the total effect model, AIP was significantly associated with increased T2DM risk (HR = 1.40, 95% CI: 1.23–1.57). When WC was included as a mediator, the natural direct effect of AIP on T2DM remained significant (HR = 0.96, 95% CI: 0.78–1.14), while the natural indirect effect through WC was also significant (HR = 0.47, 95% CI: 0.40–0.53). WC mediated 33.19% (95% CI: 26.68%−39.69%) of the total association between AIP and T2DM risk. The RCS analysis stratified by WC status further demonstrated linear dose-dependent relationships between baseline AIP and T2DM risk in both normal waist circumference (p for nonlinearity = 0.229) and abdominal obesity groups (p for nonlinearity = 0.672) (Figure S4).

Association between 5-year changes in AIP and 10-year T2DM risk

Table 3; Fig. 1(B) show the association and dose-response relationship between 5-year changes in AIP and the risk of T2DM. In the unadjusted model (Model a), compared with the stable group, participants in the decreased group had a 26% lower risk of T2DM (HRs: 0.74; 95% CIs: 0.63–0.86), whereas those in the increased group had an 18% higher risk (HRs: 1.18; 95% CIs: 1.02–1.38). After adjusting for gender, age, education, marital status, income, smoking status, alcohol consumption, physical activity, family history of diabetes, systolic blood pressure, TC and BMI, the aHRs of T2DM risk for the decreased and increased groups were 0.80(95% CIs: 0.67–0.95) and 1.20 (95% CIs: 1.00–1.40), respectively, with a significant linear trend across groups (p for trend < 0.0001). Further, each 1-SD increase in the 5-year changes in AIP was associated with a 22% increase in T2DM risk (aHR: 1.20; 95% CIs: 1.11–1.29) in the fully-adjusted model (Model d). The RCS analysis confirmed a linear dose-dependent increase of 5-year changes in AIP with T2DM risk (p for nonlinearity = 0.083).

Figure 2(B) illustrates the aHRs and 95% CIs for the risk of T2DM among subgroups stratified by gender, age, smoking status, alcohol consumption, physical activity, and family history of diabetes. The positive association between AIP and T2DM risk was observed across different age groups (< 60 years: aHRs 1.18, 95% CI 1.09–1.28; ≥60 years: aHRs 1.23, 95% CI 1.05–1.44), genders (male: aHRs 1.26, 95% CI 1.12–1.43; female: aHRs 1.17, 95% CI 1.07–1.28), and smoking status (non-smoker: aHRs 1.19, 95% CI 1.10–1.30; smoker: aHRs 1.22, 95% CI 1.05–1.41). Similar patterns were observed for alcohol consumption, physical activity, and family history of diabetes. Notably, participants who were non-drinkers (aHRs 1.22, 95% CI 1.13–1.32), had ideal physical activity levels (aHRs 1.21, 95% CI 1.12–1.31), and had no family history of diabetes (aHRs 1.20, 95% CI 1.12–1.30) showed a significant association between 5-year changes in AIP and T2DM risk. No significant multiplicative interactions were found between 5-year changes in AIP and any of the examined demographic or lifestyle factors (all p for interaction > 0.05), indicating that the association between 5-year changes in AIP and T2DM risk was generally consistent across these subgroups.

To test the robustness of our findings regarding 5-year changes in AIP, sensitivity analyses were conducted using four additional models (Models e–h) (Table S4). After excluding participants with cardiovascular disease, cancer, and renal failure in Model e, the risk increased by 23% per 1-SD increment in AIP changes (aHR: 1.23; 95% CI: 1.14–1.33). When participants with impaired fasting glucose were excluded in Model f, the risk increase was 16% (aHR: 1.16; 95% CI: 1.06–1.27). Excluding individuals with dyslipidemia in Model g, the risk increase was 12% (aHR: 1.12; 95% CI: 1.00–1.26). In Model h, excluding participants who developed T2DM within 1 year after the 5-year follow-up, the risk increase was 22% (aHR: 1.22; 95% CI: 1.13–1.31).