Abstract
Prediabetes and related complications constitute significant public health burdens globally. As an indicator closely associated with abnormal glucose metabolism and atherosclerosis, the utility of Pulse Pressure Index (PPI) as a prediabetes risk marker has not been explored. We performed a retrospective cohort analysis to investigate this putative association between PPI and prediabetes hazard. Our analysis encompassed 183,517 Chinese adults ≥ 20 years registered within the Rich Healthcare Group 2010–2016. PPI was defined as (systolic blood pressure − diastolic blood pressure)/systolic blood pressure. The relationship between PPI and prediabetes risk was assessed via Cox proportional hazards regression modeling. Non-linearity evaluations applied cubic spline fitting approaches alongside smooth curve analysis. Inflection points of PPI concerning prediabetes hazard were determined using two-piecewise Cox models. During a median follow-up of 3 years (2.17–3.96 years), new-onset prediabetes was documented in 20,607 patients (11.23%). Multivariate regression analysis showed that PPI was an independent risk factor for prediabetes, and the risk of prediabetes increased by 0.6% for every 1% increase in PPI (Hazard Ratio [HR]: 1.006, 95% Confidence Interval [CI] 1.004–1.008, P < 0.001). This association was non-significant for PPI ≤ 37.41% yet exhibited a sharp upsurge when PPI surpassed 37.41% (HR: 1.013, 95% CI 1.005–1.021, P = 0.0029). Our analysis unveils a positive, non-linear association between PPI and future prediabetes risk. Within defined PPI ranges, this relationship is negligible but dramatically elevates beyond identified thresholds.
Keywords: Prediabetes, Pulse Pressure Index, Atherosclerosis, Non-Linearity
Subject terms: Cardiology, Diseases, Endocrinology, Health care, Medical research, Risk factors
Introduction
Prediabetes represents a state of elevated but not yet diabetic blood glucose levels1. It poses substantial clinical and public health challenges, conferring heightened susceptibility for type 2 diabetes, cardiovascular disease, and other adverse outcomes2,3. Current global prediabetes prevalence is estimated at 9.1% (2021), projected to rise to 10.0% by 20453. Despite these statistics, prediabetes lacks widespread awareness as a major disease precursor, with over 80% of 96 million afflicted American adults remaining undiagnosed as of 20194.
As quantified by the pulse pressure, defined as the difference between systolic and diastolic pressures, arterial stiffness acts as an independent risk predictor for cardiovascular events and mortality among both hypertensive and normotensive populations5,6. A wealth of mechanistic evidence has associated pulse pressure with vascular endothelial impairment, affirming its links to diverse health conditions like coronary artery disease, heart failure, stroke, and cognitive decline7–13. Nevertheless, the inherent variability of pulse pressure measurements constrains its utility as a cardiovascular evaluation tool. To surmount this limitation, the pulse pressure index (PPI) derived on the theoretical basis of arterial elastic properties provides an optimized indicator. Prior findings verify that PPI overcomes pulse pressure variability, serving as a valuable prognostic measure for enhanced accuracy of cardiovascular risk assessments across clinical contexts14.
Insulin resistance, defined by impaired peripheral glucose uptake in response to endogenous or exogenous insulin, is an established prediabetes and diabetes precursor sharing common genetic and environmental antecedents with atherosclerotic vascular disease15,16. Insulin resistance is a well-established intermediary between arterial stiffness and prediabetes. Previous studies have demonstrated the association between markers of arterial stiffness—such as lipid and glucose levels—and prediabetes17–19. Therefore, understanding how PPI, as a measure of arterial stiffness, correlates with prediabetes can provide valuable insights into the pathophysiological mechanisms linking these conditions14. Nevertheless, comprehensive investigations focused on the inter-relationships of PPI with prediabetes susceptibility have hitherto remained lacking, in part constrained by the limited sample sizes and predominantly cross-sectional study designs predominant in the existing literature. Hence, the current study addressed these evidence gaps via a robustly-powered retrospective cohort analysis, leveraging a large multi-city dataset to delineate the potential associations between PPI and future prediabetes hazard.
Methods
Data source and acquisition
This work represents a secondary analysis of publicly accessible data from Chen et al.20, encompassing 211,833 Chinese individuals registered within the medical examination databases of the Rich Healthcare Group corporate, made available online via the Dryad Digital Repository (dataset accession link: https://doi.org/10.5061/dryad.ft8750v). In accordance with the Dryad terms of data usage, the analysis and citation of these anonymized data are permitted contingent on appropriate attribution. Our study complied with these requirements to undertake this retrospective investigation by mining this pre-existing database, thereby contributing to the cumulative research investigation efforts. The original data collection procedures implemented by Chen et al. had received ethical approval from the Rich Healthcare Group institutional review board, obviating requirements for additional ethical application within the present secondary analysis context. Moreover, the initial study was conducted in alignment with international ethical principles including the Declaration of Helsinki, ensuring adherence to pertinent human research regulations and guidelines.
Study design and procedures
The preliminary study population satisfying baseline inclusion criteria encompassed all individuals aged ≥ 20 years with ≥ 2 medical examination visits recorded within the Rich Healthcare database from 2010 to 2016 (n = 685,277). Standardized exclusions were then applied encompassing participants lacking baseline height or weight data (n = 103,946), absent designation of gender (n = 1), extreme baseline body mass index (BMI) values (defined as < 15 or > 55 kg/m2) (n = 152), missing baseline fasting plasma glucose level documentation (n = 31,370), inter-visit intervals < 2 years (n = 324,233), baseline diabetes mellitus diagnoses either via self-report or fasting plasma glucose ≥ 7.0 mmol/L (n = 7112), alongside undefined diabetes status at follow-up (n = 6630). Following implementation of these criteria, the final study cohort for analysis totaled to 211,833 eligible participants20.
Within this cohort, prediabetes status was delineated base on the standard definition of a fasting plasma glucose level ranging from 5.6 to 6.9 mmol/L21. Additional exclusions applied under the present study encompassed: (1) Baseline fasting plasma glucose ≥ 5.6 mmol/L (n = 26,247) (2) Follow-up fasting plasma glucose exceeding 6.9 mmol/L threshold (n = 3562) (3) Incident diagnoses of diabetes mellitus during follow-up (n = 4174), (4) Unavailable fasting plasma glucose values at follow-up examinations (n = 19), (5) Absent systolic blood pressure information (n = 23), (6) Absent diastolic blood pressure data (n = 24), and (7) Extreme outlier pulse pressure index values, defined as above or below three standard deviations from the total cohort mean (n = 954). The final sample size of 183,517 participants was obtained after additional excluding individuals with missing key variable data, baseline diabetes, and follow-up less than two years.
For participants who visited multiple times, the visit where prediabetes was first diagnosed was considered the visit of interest, otherwise the last visit was considered the study endpoint.
Data collection and measurements
The dataset from Chen et al.20 incorporated diverse demographics alongside detailed clinical and laboratory information captured during consecutive visits completed by the study participants at health screening centers affiliated with the Rich Healthcare Group corporate. Self-administered health questionnaires were utilized to collate participant demographics encompassing age, gender, lifestyle factors (e.g. smoking and alcohol consumption), medical histories (prior diagnoses), and first-degree family histories related to select chronic diseases. At each visit, anthropometric measurements were performed by experienced nurses as per standardized protocols, encompassing body weight, height, and blood pressure values. Specifically, body weight was measured with participants wearing light indoor clothing without shoes to the nearest 0.1 kg precision. Standing body height was determined to a 0.1 cm accuracy threshold. Blood pressure recordings incorporated both systolic blood pressure (SBP) and diastolic blood pressure (DBP) measurements implemented by trained staff employing regularly calibrated mercury sphygmomanometers as per established office assessment protocols.
In conjunction, fasting venous blood sampling was implemented for plasma or serum biochemical analyses encompassing alanine aminotransferase (ALT), aspartate aminotransferase (AST), blood urea nitrogen (BUN), serum creatinine (Scr), serum triglycerides (TG), total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), and glucose levels. All samples were assayed by automated clinical analyzers (Beckman 5800) calibrated per routine laboratory practices.
The baseline Pulse Pressure Index (PPI) value for each participant was derived using the standard formula: (SBP − DBP)/SBP. For subgroup comparisons, this continuous PPI variable was further categorized by quartile-based thresholds encompassing: Q1 ≤ 0.3304; 0.3304–0.3741 for Q2; 0.3741–0.4155 for Q3; and > 0.4155 defining the Q4 upper quartile. The range and diversity of parameters captured within this database offered robust assessments pertaining to participant clinical and health profiles.
Data processing
Due to the presence of substantial missing values among select variables encompassing HDL cholesterol (45.3% absent), LDL cholesterol (45.0% absent), AST (58.5%), alongside smoking (72.3% absent) and alcohol consumption status (72.3% absent), a preliminary data processing step was implemented for appropriate handling. Firstly, the HDL, LDL, and AST categorical variables were subjected to quantile-based ordinal transformations, with missing values collectively labeled as a distinct “not recorded” category. Then, under the assumption of random missing, we adopted the method of multiple interpolation, that is, we constructed five regression models with 10 iterations and took the mean value of the obtained interpolation. The deletion of TG (2.3% deletion), TC (2.3% deletion), ALT (0.8% deletion), BUN (10.1% deletion), and Scr (5.3% deletion) variables were resolved. Finally, the derived PPI values were computed and categorized into quartiles as delineated above to enable more granular analytic comparisons. By applying these evidence-based missing data handling approaches, datasets were appropriately optimized to ensure robust conclusions.
Statistical analyses
Normality assumptions were tested via Kolmogorov–Smirnov statistics. Continuous variables conforming to normal distributions were summarized as means alongside standard deviation values, with between-group statistical differences assessed using unpaired Student’s t tests. For non-normally distributed continuous variables, median and interquartile ranges were calculated, with intergroup comparisons applying Mann–Whitney U tests. Categorical variables were presented as quantities alongside percentages, employing Pearson’s chi-square or Fisher’s exact tests for intergroup comparisons as dictated by subgroup strata sizes.
Incidence rates for prediabetes were calculated using standard formulas for cumulative incidence and incidence density, with trends across comparison groups evaluated through logarithmic rank-based tests. Univariate Cox proportional hazard regression models were constructed to determine hazard ratios for prediabetes based on each demographic, clinical, and laboratory parameter. Multivariate Cox regressions were then performed to specifically delineate the association of PPI with prediabetes hazard following adjustment for pertinent covariates. The E-value metric was further applied to quantify the minimum strength of association that unmeasured confounders would require with both PPI and prediabetes to fully explain away the observed relationship in the multivariate models22,23.
Non-linear relationships were examined by incorporating PPI alongside all covariates within generalized additive models, applying integrated cubic spline smoothing functions and curve fitting approaches. Two-stage Cox hazard proportional models were further constructed to isolate potential PPI inflection thresholds indicative of alteration in the risk relationship with incident prediabetes. Sensitivity analyses were conducted via exclusion of subgroups with higher baseline prediabetes susceptibility encompassing elderly participants ≥ 75 years and obese status (BMI ≥ 25 kg/m2). Stratified subgroup analyses were also performed based on pivotal demographic and clinical parameters incorporating age, gender, BMI, family history of diabetes, alcohol intake, and smoking. In all subgroup assessments, comprehensive adjustment was made for each of the stratification variables examined. Likelihood ratio tests were applied to verify differences between strata. Statistical significance was defined by a two-sided alpha threshold of 0.05 across all analytic undertakings to ensure robust appraisal of the obtained results.
Ethics approval
The original study followed the guidelines outlined in the Declaration of Helsinki and was approved by the Rich Healthcare Group review board. In addition, the Rich Healthcare Group review committee has given up on the current informed consent of retrospective study. All methods have been implemented in accordance with the relevant Declaration of Helsinki.
Result
Participant characteristics at baseline
In total, the study population satisfying all eligibility criteria for investigation encompassed 183,517 participants (Fig. 1). Upon classification by PPI quartiles (designated as Q1–Q4 representing lowest to highest ranges), key baseline characteristic features were notable across comparison groups as compiled in Table 1. Participant mean age exhibited an inverse relationship with PPI, ranging from 36 years in the Q4 group to 39 years among Q1 participants (P < 0.001), indicative that higher PPI levels were associated with relatively younger age within this cohort. Representation of males was also higher among Q4 (55.0%) versus Q1 (52.6%, P < 0.001). In terms of metabolic and clinical parameters, participants within higher PPI quartile categories consistently exhibited significantly lower BMI, alongside SBP, DBP, fasting plasma glucose, serum triglycerides, total cholesterol, ALT, creatinine, and BUN (all P < 0.001). By comparison, the prevalence of elevated AST was slightly higher among Q4 members (14.5%) relative to Q1 participants (13.5%, P < 0.001). Rates of positive smoking history and habitual alcohol consumption both trended lower across ascending PPI quartile levels (both P < 0.001). Similarly, the prevalence of documented first-degree family history of diabetes declined from 2.2% among Q1 members to 1.7% within the Q4 group (P < 0.001).
Figure 1.
Participant selection flowchart. Extreme PPI value refers to the standard error of PPI greater than and less than 3 times the total study sample. BMI body mass index, FPG fasting plasma glucose, PPI Pulse pressure index.
Table 1.
The baseline characteristics of participants.
| Characteristic | Q1, N = 45,965 | Q2, N = 45,823 | Q3, N = 45,857 | Q4, N = 45,872 | P value |
|---|---|---|---|---|---|
| Age (years) | 39 (33, 48) | 38 (32, 47) | 37 (32, 47) | 36 (31, 50) | < 0.001 |
| Gender | < 0.001 | ||||
| Male | 24,507 (53%) | 23,875 (52%) | 23,871 (52%) | 25,090 (55%) | |
| Female | 21,458 (47%) | 21,948 (48%) | 21,986 (48%) | 20,782 (45%) | |
| BMI (kg/m2) | 22.8 (20.5, 25.2) | 22.7 (20.5, 25.0) | 22.5 (20.5, 24.8) | 22.9 (20.8, 25.1) | < 0.001 |
| SBP (mmHg) | 111 (103, 122) | 115 (105, 125) | 116 (106, 127) | 124 (113, 135) | < 0.001 |
| DBP (mmHg) | 79 (72, 87) | 74 (68, 81) | 70 (65, 77) | 68 (62, 74) | < 0.001 |
| PPI (%) | 30 (27.36, 31.71) | 35.34 (34.26, 36.43) | 39.47 (38.46, 40.52) | 44.37 (42.74, 47.18) | < 0.001 |
| FPG (mmol/L) | 4.80 (4.41, 5.10) | 4.80 (4.45, 5.10) | 4.82 (4.49, 5.12) | 4.87 (4.54, 5.16) | < 0.001 |
| TC (mmol/L) | 4.62 (4.09, 5.22) | 4.60 (4.06, 5.18) | 4.56 (4.01, 5.15) | 4.56 (4.01, 5.17) | < 0.001 |
| TG (mmol/L) | 1.07 (0.73, 1.63) | 1.03 (0.71, 1.55) | 1.00 (0.70, 1.49) | 1.01 (0.70, 1.50) | < 0.001 |
| QHDL | < 0.001 | ||||
| Not recorded | 20,378 (44%) | 20,551 (45%) | 20,961 (46%) | 21,165 (46%) | |
| Low | 9070 (20%) | 8429 (18%) | 8139 (18%) | 7848 (17%) | |
| Medium | 8309 (18%) | 8423 (18%) | 8367 (18%) | 8388 (18%) | |
| High | 8208 (18%) | 8420 (18%) | 8390 (18%) | 8471 (18%) | |
| QLDL | < 0.001 | ||||
| Not recorded | 20,255 (44%) | 20,462 (45%) | 20,853 (45%) | 20,989 (46%) | |
| Low | 8258 (18%) | 8348 (18%) | 8648 (19%) | 8401 (18%) | |
| Medium | 8555 (19%) | 8600 (19%) | 8295 (18%) | 8202 (18%) | |
| High | 8897 (19%) | 8413 (18%) | 8061 (18%) | 8280 (18%) | |
| ALT, U/L | 18 (13, 28) | 17 (12, 27) | 17 (12, 26) | 18 (13, 26) | < 0.001 |
| QAST | < 0.001 | ||||
| Not recorded | 27,285 (59%) | 27,153 (59%) | 26,775 (58%) | 26,083 (57%) | |
| Low | 5965 (13%) | 6357 (14%) | 6720 (15%) | 6366 (14%) | |
| Medium | 6245 (14%) | 6174 (13%) | 6372 (14%) | 6618 (14%) | |
| High | 6470 (14%) | 6139 (13%) | 5990 (13%) | 6805 (15%) | |
| BUN (mmol/L) | 4.48 (3.78, 5.29) | 4.47 (3.75, 5.29) | 4.49 (3.77, 5.30) | 4.56 (3.82, 5.40) | < 0.001 |
| Scr (μmol/L) | 69 (57, 80) | 68 (57, 80) | 68 (57, 80) | 70 (58, 81) | < 0.001 |
| Smoking | < 0.001 | ||||
| Not recorded | 32,495 (71%) | 33,281 (73%) | 33,360 (73%) | 33,540 (73%) | |
| Current | 2746 (6.0%) | 2487 (5.4%) | 2297 (5.0%) | 2107 (4.6%) | |
| Ever | 553 (1.2%) | 516 (1.1%) | 507 (1.1%) | 528 (1.2%) | |
| Never | 10,171 (22%) | 9539 (21%) | 9693 (21%) | 9697 (21%) | |
| Drinking | < 0.001 | ||||
| Not recorded | 32,495 (71%) | 33,281 (73%) | 33,360 (73%) | 33,540 (73%) | |
| Current | 269 (0.6%) | 273 (0.6%) | 237 (0.5%) | 205 (0.4%) | |
| Ever | 1851 (4.0%) | 1796 (3.9%) | 1857 (4.0%) | 1847 (4.0%) | |
| Never | 11,350 (25%) | 10,473 (23%) | 10,403 (23%) | 10,280 (22%) | |
| Family history of diabetes | 1015 (2.2%) | 945 (2.1%) | 902 (2.0%) | 772 (1.7%) | < 0.001 |
SBP systolic blood pressure, DBP diastolic blood pressure, BMI body mass index, ALT alanine aminotransferase, AST aspartate aminotransferase, HDL-C high-density lipoprotein cholesterol, LDL-C low-density lipoprotein cholesterol, TC total cholesterol, TG triglycerides, Scr serum creatinine, BUN blood urea nitrogen, FPG fasting plasma glucose, PPI pulse pressure index.
Prediabetes incidence stratified by PPI
During a median follow-up of 3 years (2.17–3.96 years), new-onset prediabetes was documented in 20,607 patients (11.23%). Evaluation of prediabetes incidence stratified according to PPI quartile classification demonstrated consistent dose-dependent elevations in incidence rate (P for trend < 0.001) and incidence densities (P for trend < 0.001) across the Q1 to Q4 range (Fig. 2). Kaplan–Meier survival curves examining preservation of prediabetes-free status over follow-up affirmed accelerated risk across higher PPI strata (Fig. 3). Specifically, groups Q1 and Q2 demonstrated analogous survival trajectories, whereas the Q3 and Q4 categories verified markedly increased hazard of incident prediabetes confirmation in tandem with elevated PPI measures from baseline (P-Log-Rank < 0.0001). Together these results verify that higher PPI levels incur a heightened susceptibility for developing prediabetes over prospective observation.
Figure 2.
Prediabetes incidence across PPI groups. PPI pulse pressure index.
Figure 3.
Kaplan–Meier survival curves. PPI pulse pressure index.
Evaluation of the risk factor for prediabetes
Univariate Cox proportional hazard regression modeling for associated risk of incident prediabetes conditional on each demographic, clinical, and biochemical parameter are outlined in Table 2. Advanced age conferred greater prediabetes hazard, with a 1-year increment above the mean baseline value independently predicting 3.3% greater risk (Hazard Ratio [HR]: 1.033; 95% Confidence Interval [CI] 1.032–1.034; P < 0.001). Female gender was associated with a lower hazard compared to males (HR: 0.633; 95% CI 0.615–0.651; P < 0.001). Higher BMI predicted 12.5% greater risk per unit elevation (95% CI 1.120–1.129; P < 0.001). Both SBP and DBP associated positively with prediabetes hazard (Per mmHg SBP elevation, HR: 1.025; 95% CI 1.024–1.026, P < 0.001; Per 1 mmHg DBP rise, HR: 1.030; 95% CI 1.029–1.031; P < 0.001). Relative to active smoking, never smoker status registered a 28.8% lower risk (HR: 0.712; 95% CI 0.671–0.756; P < 0.001) whereas non-documented smoking status incurred a 23.7% reduction (HR: 0.763; 95% CI 0.723–0.805; P < 0.001). Equivalent trends were observed for drinking patterns relative to habitual alcohol consumption. Dose–response elevations in prediabetes risk were apparent for higher levels of fasting glucose, triglycerides, total cholesterol, LDL cholesterol, ALT and AST liver enzymes, alongside BUN (all P < 0.001). Higher HDL cholesterol associated with lower hazard (13.8% reduction per unit increment, 95% CI 0.856–0.864; P < 0.001). First-degree family history of diabetes and serum creatinine lacked significant associations. Most prominently, each 1% incremental rise in PPI conferred a remarkable 1.4% elevation in prediabetes hazard (HR: 1.014; 95% CI 1.012–1.016; P < 0.001), denoting it as the strongest univariate predictor within this cohort.
Table 2.
The results of univariate analysis.
| Variable | B | SE | HR (95% CI) | P value |
|---|---|---|---|---|
| Age (years) | 0.033 | 0.001 | 1.033 (1.032, 1.034) | < 0.001 |
| Gender | ||||
| Male | Ref | |||
| Female | − 0.457 | 0.015 | 0.633 (0.615, 0.651) | < 0.001 |
| BMI (kg/m2) | 0.117 | 0.002 | 1.125 (1.12, 1.129) | < 0.001 |
| SBP (mmHg) | 0.025 | 0.001 | 1.025 (1.024, 1.026) | < 0.001 |
| DBP (mmHg) | 0.03 | 0.001 | 1.03 (1.029, 1.031) | < 0.001 |
| Smoking | ||||
| Current | Ref | |||
| Ever | − 0.241 | 0.066 | 0.785 (0.69, 0.895) | < 0.001 |
| Never | − 0.34 | 0.031 | 0.712 (0.671, 0.756) | < 0.001 |
| Not recorded | − 0.271 | 0.028 | 0.763 (0.723, 0.805) | < 0.001 |
| Drinking | ||||
| Current | Ref | |||
| Ever | − 0.463 | 0.085 | 0.629 (0.533, 0.742) | < 0.001 |
| Never | − 0.543 | 0.079 | 0.581 (0.497, 0.679) | < 0.001 |
| Not recorded | − 0.527 | 0.079 | 0.59 (0.506, 0.688) | < 0.001 |
| Family history of diabetes | 0.044 | 0.045 | 1.045 (0.956, 1.142) | 0.331 |
| FPG (mmol/L) | 1.745 | 0.018 | 5.724 (5.527, 5.929) | < 0.001 |
| TC (mmol/L) | 0.201 | 0.007 | 1.223 (1.206, 1.241) | < 0.001 |
| TG (mmol/L) | 0.182 | 0.003 | 1.2 (1.191, 1.208) | < 0.001 |
| HDL | ||||
| Low | Ref | |||
| Medium | 0.083 | 0.021 | 1.087 (1.042, 1.133) | < 0.001 |
| High | − 0.034 | 0.022 | 0.966 (0.925, 1.009) | 0.121 |
| Not recorded | − 0.23 | 0.018 | 0.794 (0.766, 0.823) | < 0.001 |
| LDL | ||||
| Low | Ref | |||
| Medium | 0.199 | 0.023 | 1.22 (1.166, 1.276) | < 0.001 |
| High | 0.401 | 0.022 | 1.493 (1.43, 1.56) | < 0.001 |
| Not recorded | − 0.061 | 0.02 | 0.941 (0.905, 0.98) | 0.003 |
| ALT (U/L) | 0.003 | 0.001 | 1.003 (1.003, 1.004) | < 0.001 |
| AST | ||||
| Low | Ref | |||
| Medium | 0.129 | 0.026 | 1.137 (1.08, 1.198) | < 0.001 |
| High | 0.422 | 0.025 | 1.525 (1.451, 1.602) | < 0.001 |
| Not recorded | − 0.144 | 0.022 | 0.866 (0.829, 0.904) | < 0.001 |
| BUN (mmol/L) | 0.13 | 0.005 | 1.139 (1.127, 1.151) | < 0.001 |
| Scr (μmol/L) | 0.006 | 0.001 | 1.006 (1.005, 1.006) | < 0.001 |
| PPI (%) | 0.014 | 0.001 | 1.014 (1.012, 1.016) | < 0.001 |
SBP systolic blood pressure, DBP diastolic blood pressure, BMI body mass index, ALT alanine aminotransferase, AST aspartate aminotransferase, HDL-C high-density lipoprotein cholesterol, LDL-C low-density lipoprotein cholesterol, TC total cholesterol, TG triglycerides, Scr serum creatinine, BUN blood urea nitrogen, FPG fasting plasma glucose, PPI pulse pressure index, B regression coefficient, SE standard error, HR hazard ratio, CI confidence interval, Ref reference.
PPI is the independent risk factor for prediabetes
The independent relationship between PPI and prospective prediabetes risk was further examined through constructed multivariate Cox regression models with incremental covariate adjustment (Table 3). Within the unadjusted framework (Model 1), each 1% higher PPI registered a substantial 1.4% greater prediabetes hazard (95% CI 1.012–1.016; P < 0.001). Following adjustment for age, gender, BMI, smoking, alcohol intake, and family diabetic history (Model 2), this relationship was modestly attenuated but retained statistical significance (HR per 1% PPI rise: 1.1%; 95% CI 1.007–1.015; P < 0.001). Additional correction for other metabolic parameters and renal function indices (Model 3) induced further modest attenuation, however the association remained robust (HR: 1.006; 95% CI 1.004–1.008; P < 0.001). In addition, we developed a linear mixing model to assess the relationship between PPI and changes in blood glucose in the study population. The results of our linear mixture model analysis (Table 4) showed that the increase in PPI was significantly positively correlated with the change in blood glucose, and this relationship was statistically significant across all quartile arrays. Specifically, the fixed effect was estimated at 0.052 for Q2, 0.1 for Q3, and 0.163 for Q4 compared to reference group Q1, all showing significant increases. This suggests that people with higher PPI have a higher risk of blood sugar changes, suggesting that PPI can be an important indicator of blood sugar changes. The E-value corresponded to the Model 3 hazard estimate was 1.084, substantially exceeding the observed hazard ratio of 1.006. This denotes that any residual confounding from unmeasured or unidentified variables would necessitate an association magnitude greater than the reported risk estimate to fully negate the PPI-prediabetes relationship. Conversion of PPI from a continuous variable into quartile-based categories affirmed significant elevations in prediabetes hazard across ascending PPI strata (P < 0.001). Collectively, these results strongly point towards a potential non-linear relationship between PPI and prediabetes risk.
Table 3.
Results of different COX regression models assessing the relationship between PPI and prediabetes.
| Characteristic | HR | 95% CI | P value |
|---|---|---|---|
| Model 1 | |||
| PPI | 1.014 | (1.011, 1.017) | < 0.001 |
| Q1 | Ref | P for trend < 0.001 | |
| Q2 | 0.986 | (0.948, 1.026) | 0.485 |
| Q3 | 1.066 | (1.025, 1.108) | 0.001 |
| Q4 | 1.270 | (1.223, 1.319) | < 0.001 |
| Model 2 | |||
| PPI | 1.011 | (1.007, 1.015) | < 0.001 |
| Q1 | Ref | P for trend < 0.001 | |
| Q2 | 1.017 | (0.978, 1.058) | 0.400 |
| Q3 | 1.129 | (1.086, 1.174) | < 0.001 |
| Q4 | 1.216 | (1.171, 1.262) | < 0.001 |
| Model 3 | |||
| PPI | 1.006 | (1.004, 1.008) | < 0.001 |
| Q1 | Ref | P for trend < 0.001 | |
| Q2 | 0.996 | (0.957, 1.036) | 0.833 |
| Q3 | 1.069 | (1.028, 1.112) | 0.001 |
| Q4 | 1.106 | (1.065, 1.149) | < 0.001 |
Model 1: No covariates are adjusted; Model 2: Age, gender, BMI, smoking status, drinking status, and family history of diabetes were adjusted as covariates; Model 3: Age, gender, BMI, smoking status, drinking status, family history of diabetes, TC, TG, HDL-C, LDL-C, AST, ALT, Scr, BUN and FPG were adjusted as covariates.
BMI body mass index, ALT alanine aminotransferase, AST aspartate aminotransferase, HDL-C high-density lipoprotein cholesterol, LDL-C low-density lipoprotein cholesterol, TC total cholesterol, TG triglycerides, Scr serum creatinine, BUN blood urea nitrogen, FPG fasting plasma glucose, PPI pulse pressure index, HR hazard ratio, CI confidence interval, Ref reference.
Table 4.
the result of linear mixing model.
| Variable | Estimation | SE | 95% CI | P value |
|---|---|---|---|---|
| Q1 | Ref | |||
| Q2 | 0.052 | 0.006 | (0.04, 0.063) | < 0.001 |
| Q3 | 0.101 | 0.006 | (0.089, 0.113) | < 0.001 |
| Q4 | 0.163 | 0.006 | (0.152, 0.175) | < 0.001 |
SE standard error, HR hazard ratio, CI confidence interval, Ref reference.
Non-linear association of PPI with incident prediabetes
The non-linear PPI-prediabetes risk relationship was further explored by incorporating PPI alongside adjustment covariates within a generalized additive model, applying integrated spline smoothing techniques. As evident in Fig. 4, significant non-linear patterns were consistently observed between PPI and prediabetes hazard irrespective of model adjustments (all P < 0.05). Building upon these results, a two-piecewise Cox model approach identified a clear PPI inflection threshold at 37.41% (P < 0.001), indicative of a dramatic modulatory shift in risk trajectory beyond this level (Table 5). When PPI remained below this cutoff (PPI ≤ 37.41%), no significant associative relationship with prediabetes was detectable (HR: 1; 95% CI 1–1.002; P > 0.05). However, upon surpassing this threshold (PPI > 37.41%), risk rose markedly with each 1% increment predicting a 1.3% elevation in prediabetes hazard (HR: 1.013; 95% CI 1.005–1.021; P = 0.0029). Together, these analyses confirm the presence of a significant non-linear correlation between PPI and susceptibility to developing prediabetes.
Figure 4.
Non-linear PPI-prediabetes relationship. Model 1: No covariates are adjusted; Model 2: Age, gender, BMI, smoking status, drinking status, and family history of diabetes were adjusted as covariates; Model 3: Age, gender, BMI, smoking status, drinking status, family history of diabetes, TC, TG, HDL-C, LDL-C, AST, ALT, Scr, BUN and FPG were adjusted as covariates. BMI body mass index, ALT alanine aminotransferase, AST aspartate aminotransferase, HDL-C high-density lipoprotein cholesterol, LDL-C low-density lipoprotein cholesterol, TC total cholesterol, TG triglycerides, Scr serum creatinine, BUN blood urea nitrogen, FPG fasting plasma glucose, PPI pulse pressure index.
Table 5.
The result of the piecewise Cox proportional hazards regression model.
| Variable | B | SE | HR (95%CI) | P value |
|---|---|---|---|---|
| Standard model | 0.006 | 0.001 | 1.006 (1.004, 1.008) | < 0.001 |
| Piecewise model | ||||
| PPI ≤ 37.41% | 0 | 0.00019 | 1 (1, 1.002) | > 0.9 |
| PPI > 37.41% | 0.013 | 0.004 | 1.013 (1.005, 1.021) | 0.0029 |
| P for log likelihood ratio test < 0.001 | ||||
B regression coefficient, SE standard error, HR hazard ratio, CI confidence interval, Ref reference.
Sensitivity and subgroup analyses
The robustness of the PPI-prediabetes relationship was further investigated through the sensitivity analysis of the two models, as shown in Table 6. Model 1 performed sensitivity analysis on participants < 75 years of age in the cohort, while Model 2 analyzed individuals with a BMI < 25 kg/m2. In both cases, the association between elevated PPI and prediabetes risk remained statistically significant (Model 1: HR per 1% PPI rise: 1.005; 95% CI 1.003–1.007, P < 0.001; Model 2: HR per 1% PPI increase: 1.009, 95% CI 1.007–1.011, P < 0.001). When stratifying by age and BMI categories, consistent results were observed corroborating the significant links between elevated PPI and heightened prediabetes risk. Among subgroup analyses encompassing other pivotal parameters (Table 7), higher PPI remained a significant predictor for increased prediabetes susceptibility across nearly all major subpopulations incorporating age < 75 years, female gender, non-smoking status, non-alcohol consumers, and absence of first-degree diabetes family histories (all P < 0.05). Interestingly, more pronounced PPI-prediabetes associations were noted among women (HR: 1.008), alongside persons with BMI < 25 kg/m2 (HR: 1.009) versus higher BMI (HR: 1.001). Similarly, missing documentation for smoking and alcohol intake registered stronger PPI-prediabetes links compared to measured statuses for each corresponding factor. By comparison, non-significant hazard relationships were observed among male participants, presence of overweight/obesity (BMI ≥ 25 kg/m2), smokers, alcohol consumers, and positive family history subgroups following multivariate adjustment. Collectively, these findings verify broad consistency in the association between heightened PPI and greater projected prediabetes hazard, particularly among women alongside leaner and more metabolically healthy individuals lacking detailed lifestyle histories.
Table 6.
The relationship between PPI and prediabetes in different sensitivity analyses.
| Model 1 | Model 2 | |||||||
|---|---|---|---|---|---|---|---|---|
| (Age < 75 years) | (BMI < 25 kg/m2) | |||||||
| B | SE | HR (95% CI) | P value | B | SE | HR (95% CI) | P value | |
| PPI | 0.005 | 0.001 | 1.005 (1.003, 1.007) | < 0.001 | 0.009 | 0.001 | 1.009 (1.007, 1.011) | < 0.001 |
| Category | P for trend < 0.001 | P for trend < 0.001 | ||||||
| Q1 | Ref | Ref | ||||||
| Q2 | − 0.015 | 0.02 | 0.986 (0.947, 1.025) | 0.469 | 0.01 | 0.026 | 1.01 (0.959, 1.064) | 0.708 |
| Q3 | 0.048 | 0.02 | 1.049 (1.009, 1.091) | 0.016 | 0.082 | 0.026 | 1.086 (1.032, 1.142) | 0.001 |
| Q4 | 0.094 | 0.02 | 1.098 (1.057, 1.142) | < 0.001 | 0.163 | 0.025 | 1.177 (1.12, 1.237) | < 0.001 |
B regression coefficient, SE standard error, HR hazard ratio, CI confidence interval, Ref reference.
Table 7.
The relationship between PPI and prediabetes in different subgroup analysis.
| Variable | B | SE | HR (95%CI) | P value | P for interaction |
|---|---|---|---|---|---|
| Age | 0.028 | ||||
| < 75 years | 0.005 | 0.001 | 1.005 (1.003, 1.007) | < 0.001 | |
| ≥ 75 years | 0.014 | 0.005 | 1.014 (1.003, 1.050) | 0.009 | |
| Gender | < 0.001 | ||||
| Male | 0.002 | 0.001 | 1.002 (0.999, 1.004) | 0.12 | |
| Female | 0.008 | 0.002 | 1.008 (1.004, 1.014) | < 0.001 | |
| BMI | < 0.001 | ||||
| < 25 kg/m2 | 0.009 | 0.001 | 1.009 (1.007, 1.012) | < 0.001 | |
| ≥ 25 kg/m2 | 0.001 | 0.002 | 1.001 (0.998, 1.004) | 0.661 | |
| Smoking | < 0.001 | ||||
| Current | 0.002 | 0.004 | 1.002 (0.995, 1.011) | 0.626 | |
| Ever | − 0.01 | 0.009 | 0.990 (0.943, 1.221) | 0.264 | |
| Never | − 0.001 | 0.002 | 0.999 (0.995, 1.003) | 0.558 | |
| Not recorded | 0.008 | 0.001 | 1.008 (1.006, 1.012) | < 0.001 | |
| Drinking | < 0.001 | ||||
| Current | − 0.012 | 0.013 | 0.988 (0.974, 1.036) | 0.334 | |
| Ever | − 0.001 | 0.005 | 0.999 (0.993, 1.012) | 0.876 | |
| Never | − 0.001 | 0.002 | 0.999 (0.995, 1.003) | 0.803 | |
| Not recorded | 0.008 | 0.001 | 1.008 (1.006, 1.012) | < 0.001 | |
| Family history of diabetes | 0.096 | ||||
| No | 0.006 | 0.001 | 1.006 (1.004, 1.011) | < 0.001 | |
| Yes | − 0.005 | 0.007 | 0.995 (0.988, 1.024) | 0.45 |
B regression coefficient, SE standard error, HR hazard ratio, CI confidence interval, Ref reference.
Discussion
This large retrospective cohort investigation in a real-world East Asian population align with existing literature that highlights the pathway from arterial stiffness to insulin resistance and subsequently to prediabetes. By investigating the association between PPI and prediabetes, we aimed to build on the understanding that measures of arterial stiffness are relevant markers for metabolic disturbances, including insulin resistance and prediabetes. This study reinforces the importance of arterial stiffness in the early identification of individuals at risk for metabolic disorders. This relationship was further noted to conform to a significant non-linear pattern, with negligible risk increment within lower PPI ranges followed by an exponential rise beyond an identified threshold of 37.41%. This suggests that there is a complex relationship between insulin resistance and atherosclerosis. First, insulin resistance damages endothelial cells and worsens atherosclerosis24. Secondly, damage of micro-vessels caused by arteriosclerosis may affect pancreatic function and induce insulin resistance19,25. Although this vicious circle is compensated within the body's tolerance range, the vicious circle will be broken once it exceeds this range. This also suggests that the effects of microvascular damage may increase rapidly after a certain tipping point, exacerbating insulin resistance.
Remarkably robust predictive links were affirmed specifically among women and non-obese participants in subgroup comparisons. These insights substantiate preliminary evidence supposing correlations of heightened arterial stiffness with deteriorating glycemic regulation based on smaller-scale or cross-sectional observations26,27. From a methodological perspective, our research addressed critical limitations in the existing literature centered on instability of routine pulse pressure quantifications28,29, by implementing PPI as a more reliable atherosclerosis gauge with superior risk discrimination and prognostic accuracy14. Building upon prior datasets constrained by scope, sample size, and temporal analytic approaches26,27, the current project marshaled a vast cohort to investigate these relationships from an invaluable longitudinal perspective. Our work thereby extends understanding of atherosclerotic disease processes as candidate contributors toward dysglycemia14,28,30, complementing extensive evidence implicating shared antecedents like inflammation, oxidative stress, and insulin resistance as pathogenic links between vascular and metabolic dysfunction31–38. Collectively, these merits serve to substantiate preliminary PPI evidence and provide impetus for additional epidemiological and mechanistic investigations exploring therapeutic avenues.
Beyond univariate appraisals demonstrating powerful links between PPI and prediabetes hazard, robust significance was retained across sensitivity and subgroup analyses indicating broad generalizability of this relationship independent of confounders. Our research addressed a hitherto unexplored question of outstanding uncertainty regarding whether arterial stiffness indicators may confer utility for refined phenotyping of metabolic disease risk. These results provide the first indication that PPI may distinguish subsets of individuals exhibiting heightened prediabetes susceptibility even among ostensibly low-risk cohorts lacking obvious lifestyle risk factors. Indeed, associations were particularly enhanced among women alongside persons without obesity or detailed records of smoking and alcohol behaviors. These trends were unexpected yet important observations. Despite exhibiting lower baseline hazard on aggregate analysis, women developing prediabetes registered substantially greater risk increments conditional on increasing PPI relative to men. This trend remained highly significant after excluding cardio-metabolic comorbidities and other common confounders. This suggests women may intrinsically harbor greater susceptibility toward arterial aging-related glucoregulatory impairment39,40. This may be attributed to metabolic differences between men and women due to differences in body composition41. Beyond gender, stronger PPI-prediabetes links among lean subgroups lacking lifestyle histories hint at increased dysmetabolic sensitivity to vascular injury signals across ostensibly low-risk populations42. The absence of associations among overweight/obese individuals with positive smoking or drinking histories suggests deleterious impacts from these factors may supersede subtly abnormal arterial function in driving prediabetes risk. In addition, normal-weight individuals in China may not monitor their health indicators or take preventive measures43. Nevertheless, our study represents an early foray into this research domain without precedential comparisons. Further work should probe these tendencies through detailed phenotype-wide investigation. On balance, these subgroup findings reveal new clinical avenues wherein PPI may confer advantages as a functional prediabetes marker among individuals lacking obvious risks, expediting early preventive intervention.
Our documentation of a significant non-linear relationship between PPI and prediabetes susceptibility constitutes another seminal observation laying groundwork for future analytics. Within defined boundaries up to 37.41%, PPI elevation had negligible effects on prediabetes hazard. However, exceeding this level led to an exponential 1.3% rise in risk with every 1% PPI increment. These patterns of subtle initial changes followed by sharp dynamics mirrors relationships of small and large arterial elasticity indices with metabolic traits including arterial compliance measures noted in preceding analyses44,45. Collectively, this implies potential plateau effects of buffering mechanisms up to moderate arterial stiffening thresholds. Upon exceeding compensatory limits however, risk dramatically escalates denoting failure to mitigate pulsatile stresses. Our identified PPI cut-point offers preliminary standardization for gauging transition into a high-prediabetes likelihood state on routine assessments. Additionally, this critical value provides impetus for exploring tailored interventions among individuals surpassing stipulated PPI cutoffs. Nevertheless, further confirmation across disparate samples and settings is necessary prior to translation into management algorithms. Cognizant of these considerations, our documentation of a nonlinear risk pattern remains a vital first step for instigating more focused analytics.
Limitations
Some limitations in the present study warrant due acknowledgment. As with any retrospective analysis, inherent constraints related to exclusion of unmeasured parameters that may exert residual confounding effects should be noted. However, our deliberate appraisal of an extensive panel of cardio-metabolic traits alongside application of E-value sensitivity checks substantiate confidence in the reported relationships being minimally affected by omitted factors. Additionally, external validity of findings derived from a single East Asian database may have limited extrapolation to diverse populations. Nevertheless, the size, setting diversity, and real-world nature of this cohort offers reasonable approximation of generalized risk patterns to inform hypothesis generation. Moreover, the lack of oral glucose tolerance testing represents another relative constraint precluding definitive discrimination of all prediabetes cases compared to reliance on fasting glucose thresholds. In addition, diagnostic errors caused by a single blood glucose measurement are inevitable. Therefore, large-scale clinical studies are still needed in the future to normatively explore the potential mediating role of dynamic changes in blood glucose in the relationship between atherosclerosis and prediabetes, such as PPI. Finally, the cross-sectional nature of PPI measurements provides only a snapshot representation of arterial function lacking longitudinal appraisal of time-varying changes. Hence further studies with serial imaging are warranted to clarify how PPI trajectory alterations associate with efficient prediabetes evolution.
Conclusion
In summary, this large retrospective East Asian cohort investigation performed comprehensive appraisal of putative interrelationships between the optimized atherosclerosis indicator PPI and susceptibility toward developing prediabetes. Our analyses revealed a powerful independent positive association between incremental PPI elevation and risk of prospective prediabetes onset, retaining significance across sensitivity assessments among ostensibly low-risk subgroups without documented lifestyles alongside more metabolically healthy individuals. This relationship was further noted to conform to a significant non-linear pattern, with negligible risk increment within lower PPI ranges followed by an exponential rise beyond an identified threshold of 37.41%. Remarkably robust predictive links were affirmed specifically among women and non-obese participants. These new insights substantiate preliminary suppositions based on smaller-scale or cross-sectional observations, signifying arterial stiffness indices as candidate biomarkers distinguishing subsets of individuals exhibiting heightened metabolic disease risk. Our work thereby provides impetus for additional epidemiological and mechanistic investigations focused on clarifying the impacts of vascular aging trajectories on dysglycemia to identify appropriate interventions for translation into prediabetes screening and management algorithms.
Acknowledgements
We sincerely thank Chen et al. for organizing and sharing the data. And we also thanks to the Rich Healthcare Group Review Board for data collection. The research presented in this manuscript was supported by grants from the Fundamental Research Funds for the Central Universities and the National Natural Science Foundation of China, which facilitated the execution of this study.
Abbreviations
- ALT
Alanine aminotransferase
- AST
Aspartate aminotransferase
- BMI
Body mass index
- BUN
Blood urea nitrogen
- CI
Confidence interval
- DBP
Diastolic blood pressure
- FPG
Fasting plasma glucose
- HDL-C
High-density lipoprotein cholesterol
- HR
Hazard ratio
- LDL-C
Low-density lipoprotein cholesterol
- PPI
Pulse pressure index
- SBP
Systolic blood pressure
- Scr
Serum creatinine
- TC
Total cholesterol
- TG
Triglycerides
Author contributions
YP, HM, and LG are contributed equally to this work. YP: Conceptualization and Writing – Original Draft. HM: Software usage and Analysis. LG: Resources and Investigation. BK: Visualization and Methodology. WS: Writing – Review & Editing and Supervision. HH: Writing – Review & Editing and Project Administration. Approved by all co-authors; Study participant names (and other personally identifiable information) have been removed from all text/figures/tables/images in the original study.
Funding
National Natural Science Foundation of China, Grant No. 82070330.
Data availability
The datasets generated and/or analysed during the current study are available in the Dryad Digital Repository, [Persistent Web Link to Datasets: https://doi.org/10.5061/dryad.ft8750v].
Competing interests
The authors declare no competing interests.
Footnotes
Publisher's note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
These authors contributed equally: Yucheng Pan, Hong Meng and Liang Guo.
Contributor Information
Wei Shuai, Email: sw09120@163.com.
He Huang, Email: huanghe1977@whu.edu.cn.
References
- 1.Echouffo-Tcheugui, J. B., Perreault, L., Ji, L. & Dagogo-Jack, S. Diagnosis and management of prediabetes: A review. JAMA329, 1206–1216 (2023). 10.1001/jama.2023.4063 [DOI] [PubMed] [Google Scholar]
- 2.Hostalek, U. Global epidemiology of prediabetes—Present and future perspectives. Clin. Diabetes Endocrinol.5, 5 (2019). 10.1186/s40842-019-0080-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Rooney, M. R. et al. Global prevalence of prediabetes. Diabetes Care46, 1388–1394 (2023). 10.2337/dc22-2376 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.National Diabetes Statistics Report | Diabetes | CDC. https://www.cdc.gov/diabetes/data/statistics-report/index.html (2022).
- 5.Mancusi, C. et al. Higher pulse pressure and risk for cardiovascular events in patients with essential hypertension: The Campania Salute Network. Eur. J. Prev. Cardiol.25, 235–243 (2018). 10.1177/2047487317747498 [DOI] [PubMed] [Google Scholar]
- 6.Lee, J.-W., Shin, S.-J., Kim, J. & Kang, H.-T. Higher pulse pressure is associated with increased risk of cardio-cerebrovascular disease and all-cause mortality: A Korean National Cohort Study. Am. J. Hypertens.35, 647–655 (2022). 10.1093/ajh/hpac043 [DOI] [PubMed] [Google Scholar]
- 7.Winder, N. R. et al. High pulse pressure impairs cerebral artery endothelial function in young, but not old, mice. Exp. Gerontol.173, 112101 (2023). 10.1016/j.exger.2023.112101 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Takegami, M. et al. Home-measured pulse pressure is a predictor of cardiovascular disease in type 2 diabetes: The KAMOGAWA-HBP study. Nutr. Metab. Cardiovasc. Dis.32, 2330–2337 (2022). 10.1016/j.numecd.2022.08.006 [DOI] [PubMed] [Google Scholar]
- 9.Lin, C.-H. et al. Excess pressure but not pulse wave velocity is associated with cognitive function impairment: A community-based study. J. Hypertens.40, 1776–1785 (2022). 10.1097/HJH.0000000000003217 [DOI] [PubMed] [Google Scholar]
- 10.Wei, F.-F. et al. Clinical significance of mean and pulse pressure in patients with heart failure with preserved ejection fraction. Hypertension79, 241–250 (2022). 10.1161/HYPERTENSIONAHA.121.17782 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Qiu, W., Xiao, X., Cai, A., Gao, Z. & Li, L. Pulse pressure and all-cause mortality in ischaemic heart failure patients: A prospective cohort study. Ann. Med.54, 2701–2709 (2022). 10.1080/07853890.2022.2128208 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Park, J.-H. et al. Elevated pulse pressure and recurrent hemorrhagic stroke risk in stroke with cerebral microbleeds or intracerebral hemorrhage. J. Am. Heart Assoc.11, e022317 (2022). 10.1161/JAHA.121.022317 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Wu, X. et al. Pulse pressure correlates with coronary artery calcification and risk for coronary heart disease: A study of elderly individuals in the rural region of Southwest China. Coron. Artery Dis.30, 297–302 (2019). 10.1097/MCA.0000000000000739 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Karadavut, S., Kelesoglu, S. & Elcik, D. Relationship between the progression of coronary artery disease and pulse pressure index: A cross-sectional work. Angiology74, 687–692 (2023). 10.1177/00033197221124765 [DOI] [PubMed] [Google Scholar]
- 15.Tan, L. et al. Association between insulin resistance and uncontrolled hypertension and arterial stiffness among US adults: A population-based study. Cardiovasc. Diabetol.22, 311 (2023). 10.1186/s12933-023-02038-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Mancusi, C. et al. Insulin resistance the hinge between hypertension and type 2 diabetes. High Blood Press Cardiovasc. Prev27, 515–526 (2020). 10.1007/s40292-020-00408-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Wu, J., Zhou, Q., Wei, Z., Wei, J. & Cui, M. Atherogenic index of plasma and coronary artery disease in the adult population: A meta-analysis. Front. Cardiovasc. Med.8, 817441 (2021). 10.3389/fcvm.2021.817441 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Neves, J. S. et al. Management of dyslipidemia and atherosclerotic cardiovascular risk in prediabetes. Diabetes Res. Clin. Pract.190, 109980 (2022). 10.1016/j.diabres.2022.109980 [DOI] [PubMed] [Google Scholar]
- 19.Zheng, X., Zhang, X., Han, Y., Hu, H. & Cao, C. Nonlinear relationship between atherogenic index of plasma and the risk of prediabetes: A retrospective study based on Chinese adults. Cardiovasc. Diabetol.22, 205 (2023). 10.1186/s12933-023-01934-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Chen, Y. et al. Association of body mass index and age with incident diabetes in Chinese adults: A population-based cohort study. BMJ Open8, e021768 (2018). 10.1136/bmjopen-2018-021768 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Tuomilehto, J. et al. Prevention of type 2 diabetes mellitus by changes in lifestyle among subjects with impaired glucose tolerance. N. Engl. J. Med.344, 1343–1350 (2001). 10.1056/NEJM200105033441801 [DOI] [PubMed] [Google Scholar]
- 22.Haneuse, S., VanderWeele, T. J. & Arterburn, D. Using the E-value to assess the potential effect of unmeasured confounding in observational studies. JAMA321, 602–603 (2019). 10.1001/jama.2018.21554 [DOI] [PubMed] [Google Scholar]
- 23.Vale, C. C. R., Almeida, N. K. de O. & Almeida, R. M. V. R. de. On the use of the E-value for sensitivity analysis in epidemiologic studies. Cad Saude Publica37, e00294720 (2021). [DOI] [PubMed]
- 24.Park, K. et al. Endothelial cells induced progenitors into brown fat to reduce atherosclerosis. Circ. Res.131, 168–183 (2022). 10.1161/CIRCRESAHA.121.319582 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Huang, Z., Wang, F., Xiao, X., Liu, D. & Deng, Z. Non-linear relationship between pulse pressure and the risk of prediabetes: A 5-year cohort study in Chinese adults. Sci. Rep.14, 3824 (2024). 10.1038/s41598-024-52136-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Papadakis, J. et al. Impaired fasting glucose (IFG) prevalence among hypolipidemic treatment-naïve patients with hypertension. Curr. Hypertens. Rev.18, 153–157 (2022). 10.2174/1573402118666220321121421 [DOI] [PubMed] [Google Scholar]
- 27.Feng, L. et al. Regional variation in comorbid prediabetes and diabetes and associated factors among hypertensive individuals in Rural Bangladesh, Pakistan, and Sri Lanka. J. Obes.2019, 4914158 (2019). 10.1155/2019/4914158 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Raina, R. et al. Association of pulse pressure, pulse pressure index, and ambulatory arterial stiffness index with kidney function in a cross-sectional pediatric chronic kidney disease cohort from the CKiD study. J. Clin. Hypertens. (Greenwich)22, 1059–1069 (2020). 10.1111/jch.13905 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Boesby, L., Thijs, L., Elung-Jensen, T., Strandgaard, S. & Kamper, A.-L. Ambulatory arterial stiffness index in chronic kidney disease stage 2–5. Reproducibility and relationship with pulse wave parameters and kidney function. Scand. J. Clin. Lab. Invest. 72, 304–312 (2012). [DOI] [PubMed]
- 30.Cai, A. et al. Relationship of pulse pressure index and carotid intima-media thickness in hypertensive adults. Clin. Exp. Hypertens.37, 267–270 (2015). 10.3109/10641963.2014.954713 [DOI] [PubMed] [Google Scholar]
- 31.Qin, Z. et al. The atherogenic index of plasma plays an important role in predicting the prognosis of type 2 diabetic subjects undergoing percutaneous coronary intervention: Results from an observational cohort study in China. Cardiovasc. Diabetol.19, 23 (2020). 10.1186/s12933-020-0989-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Sonta, T. et al. Evidence for contribution of vascular NAD(P)H oxidase to increased oxidative stress in animal models of diabetes and obesity. Free Radic. Biol. Med.37, 115–123 (2004). 10.1016/j.freeradbiomed.2004.04.001 [DOI] [PubMed] [Google Scholar]
- 33.Loutzenhiser, R., Bidani, A. & Chilton, L. Renal myogenic response: Kinetic attributes and physiological role. Circ. Res.90, 1316–1324 (2002). 10.1161/01.RES.0000024262.11534.18 [DOI] [PubMed] [Google Scholar]
- 34.Satchell, S. C. & Tooke, J. E. What is the mechanism of microalbuminuria in diabetes: A role for the glomerular endothelium?. Diabetologia51, 714–725 (2008). 10.1007/s00125-008-0961-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Ma, Y. et al. Relation between blood pressure and pulse wave velocity for human arteries. Proc. Natl. Acad. Sci. USA115, 11144–11149 (2018). 10.1073/pnas.1814392115 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Milan, A. et al. Current assessment of pulse wave velocity: Comprehensive review of validation studies. J. Hypertens.37, 1547–1557 (2019). 10.1097/HJH.0000000000002081 [DOI] [PubMed] [Google Scholar]
- 37.Aminuddin, A. et al. The association between inflammation and pulse wave velocity in dyslipidemia: An evidence-based review. Mediators Inflamm.2020, 4732987 (2020). 10.1155/2020/4732987 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Liang, Y. et al. The mechanisms of the development of atherosclerosis in prediabetes. Int. J. Mol. Sci.22, 4108 (2021). 10.3390/ijms22084108 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Moreau, K. L., Clayton, Z. S., DuBose, L. E., Rosenberry, R. & Seals, D. R. Effects of regular exercise on vascular function with aging: Does sex matter?. Am. J. Physiol. Heart Circ. Physiol.326, H123–H137 (2024). 10.1152/ajpheart.00392.2023 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Zhao, X. et al. The prevalence and clustering of metabolic syndrome risk components in Chinese population: A cross-sectional study. Front. Endocrinol. (Lausanne). 14, 1290855 (2023). [DOI] [PMC free article] [PubMed]
- 41.Zeng, Y. et al. Sex differences in genetic associations with longevity. JAMA Netw. Open1, e181670 (2018). 10.1001/jamanetworkopen.2018.1670 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Bjornstad, P. et al. Insulin secretion, sensitivity, and kidney function in young individuals with type 2 diabetes. Diabetes Care. dc231818 (2023) 10.2337/dc23-1818. [DOI] [PMC free article] [PubMed]
- 43.Zhang, M. et al. Cumulative increased risk of incident type 2 diabetes mellitus with increasing triglyceride glucose index in normal-weight people: The Rural Chinese Cohort Study. Cardiovasc. Diabetol.16, 30 (2017). 10.1186/s12933-017-0514-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Kim, E. D. et al. Short-term prognostic impact of arterial stiffness in older adults without prevalent cardiovascular disease. Hypertension74, 1373–1382 (2019). 10.1161/HYPERTENSIONAHA.119.13496 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Song, B. et al. A U-shaped association between the triglyceride to high-density lipoprotein cholesterol ratio and the risk of incident type 2 diabetes mellitus in Japanese men with normal glycemic levels: A population-based longitudinal cohort study. Front. Endocrinol. (Lausanne)14, 1180910 (2023). [DOI] [PMC free article] [PubMed]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The datasets generated and/or analysed during the current study are available in the Dryad Digital Repository, [Persistent Web Link to Datasets: https://doi.org/10.5061/dryad.ft8750v].