The sample for the current research is derived from the China Hypertension Registry Study (Registration number: ChiCTR1800017274), an ongoing observational registration study conducted in Wuyuan, Jiangxi province of China during the period from March 2018 to August 2018. The study protocols have been previously described (25). The study aimed to investigate the incidence of hypertension, the condition of hypertension control, and potential risk factors affecting the prognosis of hypertension in a real-world setting. The study’s inclusion criteria comprised of individuals aged 18 years or older who had hypertension. Hypertension was defined as having a sitting systolic blood pressure (SBP) of ≥ 140 mmHg or a diastolic blood pressure (DBP) of ≥ 90 mmHg, self-reported history of hypertension, or current use of antihypertensive medication (26). The exclusion criteria comprised of the following: 1) individuals with psychological disorders or nervous system impairments that hindered their ability to comprehend informed consent; 2) participants who were unable to comply with the study protocol’s long-term follow-up requirements or had plans to relocate in the near future; or 3) individuals deemed unsuitable for inclusion or long-term follow-up by study physicians. The Ethics Committee of the Institute of Biomedicine, Anhui Medical University (NO. CH1059) and the Second Affiliated Hospital of Nanchang University (NO. 2018019) granted approval for the study protocol. Informed consent was obtained from all participants who were recruited for this study.

A hypertension cohort was formed by consecutively enrolling 14,268 adults. Subsequently, 34 non-hypertensive individuals, 5,976 individuals with an abnormal weight according to the World Health Organization (WHO) standards (BMI < 18.5, n = 975; BMI ≥ 25, n = 4,720), and 261 individuals taking lipid-lowering drugs were excluded. Ultimately, 8,258 normal-weight adults with hypertension were analyzed ( Supplementary Figure S1 ).

The study obtained demographic characteristics through questionnaire surveys administered by uniformly trained staff. The questionnaire content encompassed sex, age, smoking history, drinking history, medical history, and medication information. Anthropometric parameters included height, weight, waist circumference (WC), SBP, and DBP. Height and weight were measured without shoes, while WC was assessed at the end of expiration in an erect position using a cloth tape. The blood pressure of all participants was assessed using a consistent electronic sphygmomanometer (HBP-1300; Omron, Japan) while seated and after a minimum of 5 minutes of rest. The mean blood pressure of the right arm was measured thrice for each participant, with a 1-minute interval between each measurement, and was utilized in the final analysis. The fasting venous blood samples of all participants were collected by uniformly trained nurses and subsequently transferred to Biaojia Biotechnology in Shenzhen, China for laboratory analyses utilizing automatic clinical analyzers (Beckman Coulter, USA). The biochemical data encompassed measurements of plasma total cholesterol (TC), TG, HDL-C, LDL-C, and homocysteine (Hcy), as well as serum aspartate aminotransferase (AST), alanine aminotransferase (ALT), uric acid (UA), albumin, total bilirubin (TBiL), direct bilirubin (DBiL), and creatinine. The estimated glomerular filtration rate (eGFR) was determined using the Chronic Kidney Disease Epidemiological Collaboration equation (27).

T2DM was characterized by a fasting glucose level of ≥ 7 mmol/L, a self-reported history of diabetes, or the use of hypoglycemic therapy (28). The definition of the Atherogenic Index of Plasma (AIP) was derived from the base-10 logarithm of the TG/HDL-C ratio (29). Hyperuricemia was diagnosed as serum UA levels > 420 μmol/L (7 mg/dL), regardless of sex (30).

The baseline characteristics of the study participants were presented as mean ± standard deviation (SD) for continuous variables and number (%) for categorical variables. The data were analyzed via ANOVA for continuous variables and Chi-square test for categorical variables to identify discerning any disparities between groups based on AIP quartiles. The distribution of AIP values ​​of the subjects was normal. Multivariate logistic regression analyses were performed to assess the association between AIP and T2DM (odds ratio [OR] and 95% confidence interval [CI]). P values for the trends of groups by AIP quartiles were calculated to confirm the existence of a linear correlation between AIP and T2DM. Fourmodels were constructed to examine the correlation between AIP and T2DM under varying states of uric acid levels: crude model, unadjusted; model 1, adjusted for age, sex, WC, SBP, DBP; model 2, adjusted for age, sex, WC, SBP, DBP, current smoking, current drinking, education level, physical activity, Hcy, TC, LDL-C, AST, ALT, TBiL, DBiL, albumin, eGFR, stroke, and antihypertensive drugs; Model 3 was adjusted for age, sex, WC, BMI, SBP, DBP, current smoking, current drinking, education level, physical activity, Hcy, TC, LDL-C, AST, ALT, TBiL, DBiL, albumin, eGFR, stroke, and antihypertensive drugs.

The selection of covariates for adjustment was predicated on their clinical significance, univariable analysis, and their capacity to modify the OR values of T2DM by a minimum threshold of 10%. Additionally, a Generalized Additive Model (GAM) coupled with a penalized spline method was employed to construct a smooth fitting curve, thereby facilitating a visual evaluation of the dose-response correlation between AIP and T2DM. Upon detection of nonlinearity, a recursive algorithm was initially employed to compute the inflection point. Subsequently, a two-stage logistic regression model was constructed on either side of the inflection point. The log-likelihood ratio served as a determinant for identifying the most appropriate model to elucidate the association between AIP and T2DM. A sequence of sensitivity analyses was conducted to incrementally ascertain the reliability of the conclusions. Given the potential influence of gender and age on the prevalence of T2DM, additional multiple logistic regression analyses were executed to discern the correlation between AIP and T2DM, stratified by gender and age. In addition, E-values were calculated to assess for the effect of unmeasured confounders on the association between AIP and T2DM. Subsequently, stratified analyses were conducted to explore the potential variables that could modified the association between AIP and T2DM.

The statistical analyses were carried out utilizing R version 4.2.0 (http://www.r-project.org) and EmpowerStats (www.empowerstats.net; X&Y Solutions, Inc. Boston, Massachusetts). A statistically significant difference was determined by a two-sided P-value < 0.05.

The present study comprised a cohort of 8,258 normal-weight adults with hypertension. The average age was 64.89 ± 8.97 years and 48.60% of participants were male. The average baseline value of the AIP index was -0.04 ± 0.29. The overall prevalence of T2DM in the cohort population was 15.55%. Table 1 presents the baseline characteristics of the participants, categorized according to quartiles of AIP. There was a progressive escalation in the prevalence of T2DM as the AIP increased, with respective percentages of 10.78% (221), 12.76% (251), 14.78% (305), and 24.55% (507). In comparison to the lower AIP group, the higher AIP group demonstrated increased levels of WC, BMI, TC, TG, LDL-C, ALT, UA, and Albumin. Furthermore, the higher AIP group displayed a higher prevalence of stroke and a superior level of educational achievement. Conversely, the group with lower AIP exhibited higher values for age, Hcy, TBiL, DBiL, HDL-C, and eGFR when compared to the higher AIP group. Concurrently, the lower AIP group had a higher proportion of males, smoking, alcohol consumption, and high physical activity. Notably, there was no significant difference in systolic blood pressure and antihypertensive medication between the four groups. Eventually, the baseline characteristics of the participants were presented in Supplementary Table S1 , stratified by uric acid (UA) concentration.

Overall, there was a positive correlation between AIP and T2DM irrespective of the adjustment for confounders. The findings of the multivariate logistical regression analysis elucidating the relationship between AIP and T2DM are presented in Table 2 . Model 3 represents the fully adjusted model, which accounted for a plenty of confounders such as age, sex, WC, BMI, SBP, DBP, current smoking and drinking habits, education level, physical activity, Hcy, TC, LDL-C, AST, ALT, TBiL, DBiL, Albumin, eGFR, stroke, and usage of antihypertensive drugs. In the fully adjusted model, the odds ratio (OR) for T2DM was determined to be 3.73 (95% CI: 2.82, 4.94). When AIP was transformed into categorical variables based on quartile, the adjusted ORs for T2DM in the quartiles 2, quartiles 3, and quartiles 4 were 0.93 (95% CI: 0.74, 1.18), 1.08 (95% CI: 0.85, 1.37), and 1.99 (95% CI: 1.57, 2.52), respectively, in comparison to the quartiles 1 in Model 3 (P for trend < 0.001). Table 3 elucidates the correlation between AIP and T2DM, stratified by hyperuricemia. In the fully adjusted model, per unit increment of AIP corresponds to a 5.36-fold increase in T2DM risk among the hyperuricemia population. Similarly, AIP was divided into quartiles for further analysis. Using the quartile 1 as the reference group, the quartiles 2 and 3 did not exhibit a significant increase in T2DM risk, with ORs of 1.02 (95% CI: 0.72, 1.44) and 1.31 (95% CI: 0.93, 1.85), respectively. However, there was a significantly higher prevalence of T2DM in the quartile 4 (OR: 2.79; 95% CI: 1.98, 3.94). Conversely, within the population exhibiting normal uric acid levels, the impact of AIP on T2DM is comparatively mild (OR: 2.63; 95% CI: 1.77, 3.90). A 42% heightened risk of T2DM was observed in quartiles 4 relative to the quartiles 1. Noteworthily, there was no significant correlation between AIP and T2DM in the quartiles 2 and quartiles 3, exhibiting ORs of 0.87 (95% CI: 0.63, 1.20) and 0.88 (95% CI: 0.64, 1.21) respectively. The aforementioned findings imply a potential interaction between hyperuricemia and AIP (P for interaction = 0.034).

To verify the robustness of the findings, several sensitivity analyses were conducted. Despite the fact that this study adjusted for a substantial number of confounding variables, there remains the potential for unmeasured confounders that could affect the association between AIP and T2DM. Consequently, an E-value was computed to evaluate the potential influence of these unmeasured confounders on the study’s results. The E-value (3.29) derived from this study notably exceeded the relative risk (2.33) correlated with AIP and unmeasured confounders. This implies that the influence of unmeasured confounders on the relationship between AIP and T2DM is not sufficient to generate a false-positive outcome.

Considering that age and gender could potentially influence the relationship between AIP and T2DM, we conducted sensitivity analyses, which were stratified by age ( Supplementary Table S2 ). Multiple logistic regression analyses indicated a positive association between AIP and the risk of T2DM in both the group aged < 65 years (OR: 3.37; 95% CI: 2.26, 5.01) and the group aged ≥ 65 years (OR: 4.54; 95% CI: 3.05, 6.75). Similarly, multivariate logistic regression analyses were implemented to examine the correlation between AIP and T2DM, stratified by gender ( Supplementary Table S3 ). The findings revealed that the association between AIP and T2DM persisted consistently in both male (OR: 4.74; 95% CI: 3.00, 7.50) and female (OR: 3.20; 95% CI: 2.26, 4.55).

Generalized additive models (GAM) and fitted smoothing curves (penalized spline method) were performed to visually assess the correlation between AIP and the risk of T2DM. After adjusting for relevant confounders, a J-shaped association was detected between AIP and T2DM, with a inflection point of 0 identified by recursive algorithm ( Figure 1 ; Table 4 ). when AIP was > 0, AIP was independently and positively associated with the risk of T2DM (OR: 6.20; 95% CI: 4.20, 9.14). conversely, when AIP was < 0, there was no significant correlation between AIP and T2DM (OR: 1.37; 95% CI: 0.75, 2.51).

Stratified analyses were conducted to detect potential variables that could modified the relationship between AIP and the risk of T2DM ( Figure 2 ) The correlation between AIP and T2DM remained consistent across the following subgroups: age (< 65 vs. ≥ 65 year), sex (male vs. ≥ female), physical activity (mild, moderate, vigorous), current smoking status (no vs. yes), drinking status (no vs. yes), TC (< 5.20 vs. ≥ 5.20 mmol/L), and LDL-C (< 2.6 vs. ≥ 2.6 mmol/L). Nevertheless, the association between AIP and T2DM was significantly stronger in the hyperuricemia population compared to individuals with normal uric acid levels (P for interaction = 0.034).

It is essential to recognize both the strengths and limitations of the present study when interpreting our findings. To our knowledge, the current study is the first to investigate the relationship between AIP and the risk of T2DM in normal-weight adults with hypertension. Furthermore, in assessing the independent association between AIP and T2DM, this study accounted for a substantial number of confounding variables and evaluated the potential influence of unknown confounders by calculating the E value. This methodological approach was employed to mitigate the impact of confounding factors on the results to the greatest extent possible. Additionally, stratified analyses were conducted to assess the robustness of the findings and to identify potential effect modifiers that may influence the correlation between AIP and T2DM. Nonetheless, it is imperative to acknowledge several limitations inherent in this study. Primarily, the cross-sectional design was insufficient to establish a causality between AIP and T2DM. Secondly, the diagnostic criterion utilized in this study was fasting plasma glucose. However, measurements of glycosylated hemoglobin (HbA1c) levels were not undertaken, nor were oral glucose tolerance tests (OGTT) administered. Consequently, there is a potential for diagnostic misclassification, whereby some diabetic patients may not have been accurately identified. This limitation could lead to an underestimation of the association between AIP and T2DM. It is noteworthy that the Framingham Heart Study also employed fasting blood glucose levels as the method for diagnosing T2DM (54). Thirdly, although the current study adjusted for a substantial number of confounding variables to mitigate their impact on the research outcomes, the inherent limitations of cross-sectional studies cannot entirely preclude the influence of unmeasured confounding factors. Nevertheless, through the calculation of the E-value, it was determined that the impact of these unknown confounders on the study’s results is negligible. Finally, the study’s sample comprised normal-weight Chinese adults with hypertension, which restricts the generalizability of the findings to other populations.