The NHANES uses a multistage, stratified, and clustered sampling design to collect nationally representative health examination data from the US population, including basic information on participants, laboratory test results, and underlying diseases. The NHANES protocol was approved by the Ethics Review Committee of the National Center for Health Statistics (NCHS), and written consent was obtained from each participant prior to their involvement in the study. The research procedures were authorized by the NCHS Ethics Review Board, and the study complied with the ethical standards set forth in the Declaration of Helsinki. For this cross-sectional study, we used data from five cycles of the NHANES (2009-2018), that included a cohort of 49,694 participants. Our study excluded individuals under 20 years of age (n = 21,357), participants without a diagnosis of hypertension (n = 17,424), those with missing data on high-density lipoprotein cholesterol or total cholesterol (TC) levels, those lacking diagnostic criteria for diabetes or prediabetes, and pregnant or lactating females. After these exclusions, 10,250 participants were included in the final analysis ( Figure 1 ).

Sample collection followed strict standardized procedures conducted at Mobile Examination Centers. Following venous blood collection, samples were transported to the central laboratory for analysis. If testing could not be conducted within the designated time frame, samples were either refrigerated or frozen at appropriate temperatures. All experimental procedures were performed by trained professionals in accordance with standardized protocols (19).

Hypertension was assessed by averaging three–two blood pressure measurements. If only one measurement was available, the use of antihypertensive medication was also considered. Individuals were classified as hypertensive if their systolic blood pressure was ≥130 mmHg and/or diastolic blood pressure was ≥80 mmHg. Blood pressure measurements are conducted by trained technicians using either mercury sphygmomanometers or electronic blood pressure devices, following established international guidelines (20).

The exposure variable, NHHR, was calculated as non-HDL-C divided by HDL-C (21). Non-HDL-C was determined by subtracting HDL-C from TC. HDL-C and TC concentrations were measured using precipitation or immunoassay techniques. Participants were divided into four groups (Q1, Q2, Q3, and Q4) based on NHHR quartiles, with Q1 serving as the reference group.

According to international guidelines and prior studies (22), diabetes was diagnosed based on: 1) physician-conformed diagnosis; 2) fasting blood glucose (FPG) levels ≥7.0 mmol/L; 3) hemoglobin A1c (HbA1c) ≥6.5%; 4) use of anti-diabetic medication or insulin. Prediabetes was defined as (3): self-reported prediabetes, 5.7% ≤ HbA1c <6.5%, or FPG levels ranged from 5.6 to 7.0 mmol/L.

Covariate selection was based on existing literature and clinical expertise, considering multiple confounders that may affect NHHR and diabetes risk. Demographic variables included age, sex, ethnicity, educational level, marital status, family income-to-poverty ratio (PIR), and body mass index (BMI; 18.5–25 kg/m², 25–30 kg/m², ≥30 kg/m²). Lifestyle variables included smoking status (never, former, current), alcohol consumption (yes/no), and physical activity (PA). Alcohol use was defined as consumption of at least 12 alcoholic beverages annually. PA was categorized using total metabolic equivalent (MET) minutes per week, following U.S. Physical Activity Guidelines: inactive (0 MET-min/week), insufficiently active (<600 MET-min/week), and active (>600 MET-min/week). Laboratory measures included albumin, HbA1c, total cholesterol (TC), triglyceride (TG), HDL-C, serum creatinine (SCR), and uric acid. Comorbidities included cardiovascular disease (CVD) and chronic kidney disease (CKD), with CKD defined as self-reported CKD, eGFR <60 ml/min/1.73 m², or urine albumin-to-creatinine ratio (UACR) ≥30 mg/g.

Statistical analyses were performed using R software (version 4.1.0) and Free Statistics software. NHANES weights were applied according to guidelines, with data spanning five cycles (2009-2018) and adjusted using WTMEC2YR/5. Continuous variables are presented as means ± standard deviations (SD), while categorical variables are expressed as percentages. Group differences by NHHR quartile were assessed using chi-square tests for categorical variables, Kruskal–Wallis test for non-normally distributed continuous variables, and t-tests for normally distributed variables. Weighted logistic regression models were used to estimate odds ratios (OR) and 95% confidence intervals (CI) for the association between NHHR and diabetes/prediabetes risk. Model 1 was unadjusted; Model 2 adjusted for demographic variables (sex, age, education, race, marital status, and PIR); and Model 3 included additional adjustments for BMI, waist circumference, smoking, alcohol use, PA, uric acid, CKD, and CVD. Restricted cubic spline (RCS) curves and threshold effect analyses, based on Model 3 covariates, were used to evaluate potential nonlinear associations between NHHR and diabetes/prediabetes in patients with hypertension. Stratified analysis were conducted by age, sex, race, BMI, CKD, CVD, smoking status, alcohol use, and PA. ROC curves evaluated NHHR’s predictive value in identifying diabetes and prediabetes in this population. To enhance statistical efficiency and minimize bias, multiple imputations was applied to handle missing data. Sensitivity analysis were conducted to validate the stability of the results, revealing no significant differences between the imputed and original datasets. A two-tailed P-value less than 0.05 was considered statistically significant.

Our study included 10,250 patients with hypertension, comprising 5,561 males and 4,689 females, with a mean age of 56.31 ± 16.06 years. The overall prevalence of diabetes and prediabetes was 61.81%, with 2,198 and 4,138 individuals having diabetes and prediabetes, respectively. The participants were categorized into four groups according to NHHR quartiles: Q1 ≤ 1.98, 1.98 < Q2 ≤ 2.79, 2.79 < Q3 ≤ 3.79, and 3.79 < Q4. Tables 1 and 2 describe the baseline characteristics of the participants based on NHHR quartile and disease status, respectively. The mean (SD) of NHHR was markedly elevated in individuals with diabetes and prediabetes, at 3.14 (1.51), compared to a mean (SD) of 2.88 (1.43) in those without prediabetes. Participants with a history of diabetes or prediabetes were generally older. In terms of lifestyle, the diabetes and prediabetes groups had higher BMI and waist circumference, as well as less frequent PA. Additionally, this group exhibited more than double the prevalence of CKD and CHD compared with that of the non-prediabetes group. As the NHHR quartiles increased, the proportion of individuals with diabetes and prediabetes increased significantly (from 55.78% to 67.43%). However, there was no significant sex differences in disease prevalence between males and females.

Univariate logistic regression analysis revealed increased risk of diabetes and prediabetes with higher NHHR and non-HDL-C levels (NHHR: OR 1.13; 95% CI, 1.10-1.17; non-HDL-C: OR 1.05; 95% CI, 1.02-1.09). Individuals with a PIR > 3 had a lower risk of both the conditions than those with a PIR < 1. In addition, age, BMI, waist circumference, uric acid, triglycerides, CKD, and CVD were positively associated with the occurrence of diabetes and prediabetes, whereas HDL-C level, sufficient PA, and PIR were negatively correlated with both the conditions in this population ( Table 3 ).

As shown in Table 4 , we assessed the correlation between NHHR and diabetes/prediabetes in the hypertensive population using three logistic regression models, with the effect size represented by the OR and 95%CI. In the crude model, every unit increase in NHHR correlated with a 13% increase in the risk of diabetes/prediabetes (OR 1.13, 95% CI 1.10–1.17, P < 0.001). After analyzing NHHR as a continuous variable and adjusting for multiple potential confounders (Model 3), the multivariate logistic regression model produced results that were not significantly different from those of the unadjusted model (OR 1.21; 95% CI, 1.15-1.25, P < 0.001). Furthermore, when we converted NHHR into categorical variables based on quartiles, univariate analyses showed that individuals in the highest quartile had a significantly higher likelihood of diabetes and prediabetes compared to those in the lowest quartile (Q4 OR, 1.64; 95%CI, 1.46-1.84, P < 0.001). Adjustments performed using multivariate models revealed a 37% and 82% increase in the prevalence of diabetes and prediabetes in Q3 (OR 1.37; 95% CI, 1.27-1.59, P < 0.001) and Q4 (OR 1.82; 95% CI, 1.62-1.98, P < 0.001), using Q1 as the reference group. Thus, NHHR may serve as a potential predictor of diabetes and prediabetes in this population. Tables 5 presents the association between NHHR and the risk of diabetes and prediabetes in both male and female patients. The data revealed that as NHHR increases, the risk of developing diabetes/prediabetes is significantly higher in women compared to men (Female: OR 1.32; 95% CI, 1.23-1.41; Male: OR 1.16; 95% CI, 1.11-1.23).

Using RCS analysis, we further explored the correlation between NHHR and diabetes/prediabetes in a hypertensive population stratified by sex. After adjusting for relevant covariates, the RCS analysis revealed a nonlinear association between the NHHR and diabetes/prediabetes (nonlinear P = 0.007). Threshold effect analysis identified a cutoff point at NHHR = 7.09. Below this threshold, the OR was 1.34 (95% CI: 1.28–1.39, P < 0.001), indicating a strong positive association ( Table 6 ). In the sex-stratified analysis, a significant nonlinear relationship was observed in males (nonlinear P = 0.005), while the association appeared linear among females (nonlinear P = 0.997) ( Figure 2 ).

Stratified analysis revealed that the correlation between NHHR and diabetes/prediabetes remained consistent across most subgroups, including age, race, BMI, CKD, CVD, smoking status, alcohol consumption, and PA. However, sex significantly modified the correlation between NHHR and diabetes/prediabetes in the hypertensive population, indicating a potential sex difference. Notably, the positive association between NHHR and the occurrence of diabetes and prediabetes was stronger in the female population compared to in male population ( Figure 3 ). In the sensitivity analysis, after excluding individuals with missing baseline covariates, 7,103 individuals remained. Following adjustment for potential confounding factors, the relationship between NHHR and both diabetes and prediabetes remained consistent with the main analysis ( Table 7 ).

Figure 4 presents the ROC curve of the NHHR for predicting diabetes and prediabetes. The area under the curve (AUC) for NHHR was 0.592, with corresponding sensitivity and specificity values of 0.583 and 0.553 for diabetes and prediabetes, respectively. The area under the curve (AUC) for NHHR was 0.592, with corresponding sensitivity and specificity values of 0.583 and 0.553 for diabetes and prediabetes, respectively ( Table 8 ).

This study has several strengths. It utilized national-level data and offered a large sample size and detailed participant information, strengthening the representativeness and stability of the findings. Additionally, we adjusted for confounding factors to provide the best evidence for an association between NHHR and diabetes/prediabetes. Weighting adjustments were applied according to the NHANES analytical guidelines, and sensitivity analyses were performed with results consistent with the primary analysis, thereby increasing the credibility of the study. However, this study has some limitations. First, the NHANES database includes only baseline information at the time of TC and HDL-C measurement, which may not capture dynamic changes over time. Second, this was a retrospective study; therefore, a causal relationship between the NHHR and diabetes/prediabetes could not be established. Although we adjusted for many known confounders, the influence of unmeasured variables, such as dietary habits and stress levels, cannot be excluded. Therefore, further randomized controlled trials and longitudinal studies are warranted to validate our findings.

Our study established a significant and independent correlation between NHHR and diabetes/prediabetes risk in adults with hypertension. Specifically, a non-linear positive relationship was observed. Notably, this association was stronger in women than in men. Our study provides a basis and valuable reference for further investigations into the associations between NHHR and diabetes/prediabetes risk. Thus, NHHR may play a significant role in the early detection and management of populations at risk for both the conditions.