Abstract
Background
Prior studies on the relationship between high-density lipoprotein cholesterol (HDL-C) and depression have reported inconsistent results. Insulin resistance (IR) can alter the composition and function of HDL. This study aims to investigate whether IR influences the association between HDL-C and depression.
Methods
Data from the National Health and Nutrition Examination Survey (2005–2018) were analyzed. Depression was assessed using the Patient Health Questionnaire-9, with a score of ≥ 10 indicating depression. IR was defined by a HOMA2-IR value of ≥ 2.5. Survey-weighted generalized linear models (GLMs) were used to examine the associations between HDL-C, IR, and depression. Multiplicative and additive interaction, along with subgroup analyses, evaluated HDL-C/IR interactions affecting depression. Sensitivity analyses were conducted by: (1) redefining IR, (2) adjusting for total cholesterol and triglycerides in the base models, (3) applying alternative weighting, and (4) including special participants.
Results
The study included 7,779 participants. Survey-weighted GLMs revealed no significant association between HDL-C or IR and depression. However, HDL-C and IR had a significant synergistic effect on the odds of depression (multiplicative scale[P < 0.05] and additive interactions [relative excess risk due to interaction = 3.21]). Subgroup analyses confirmed IR significantly modified HDL-C-depression associations (Pinteraction = 0.024). Specifically, in IR-positive individuals, Higher HDL-C linked to increased depression odds (odds ratio = 4.66, 95% confidence interval: 1.23–17.59, P = 0.02).
Conclusion
Elevated HDL-C in the context of IR were associated with increased depression odds. These findings underscore the importance of considering IR when examining the relationship between HDL-C and neuropsychiatric disorders.
Keywords: Depression; Cholesterol, HDL; Insulin resistance; Synergy
Background
Depression has emerged as a critical global public health challenge in the 21 st century. According to the latest World Health Organization (WHO) report, depressive disorders affect approximately 280 million people globally and are projected to become the leading cause of worldwide disease burden by 2030 [1]. In 2019, depressive disorders accounted for 37.3% (95% uncertainty interval: 32.3–43.0) of global mental disorder-related disability-adjusted life years; these disorders ranked second among the top 25 disease categories for years lived with a disability [2]. Despite widespread clinical use of antidepressants, approximately one-third of patients exhibit suboptimal treatment responses [3]. This treatment gap necessitates moving beyond conventional neurotransmitter theories to systematically investigating novel modifiable risk factors for developing innovative interventions.
High-density lipoproteins (HDLs) comprise approximately 25–30% of plasma proteins involved in systemic lipid transport [4]. HDLs facilitate cholesterol clearance through reverse cholesterol transport and exhibit anti-inflammatory and anti-atherogenic properties [5]. Although depression is well-established to associate with inflammatory markers [6], studies on HDL-cholesterol (HDL-C) and depressive symptoms have revealed contradictory findings. Current evidence shows inconsistent associations, with reports of elevated, neutral, and reduced HDL-C levels in cases of depression [7–9].
Theoretically, HDL-mediated anti-inflammatory effects may attenuate the risk of depression. However, HDL-C levels do not fully represent HDL functionality. Pathological states can modify HDL particle composition and function [10]. Therefore, pathological factors influencing HDL biology require consideration when assessing the HDL-C–depression association. Insulin resistance (IR) significantly alters HDL composition and function [11, 12]. However, whether IR modulates the HDL-C–depression association remains uninvestigated.
Using National Health and Nutrition Examination Survey (NHANES) data, this study examines the HDL-C-depression association. First, individuals using antidepressants or lipid-lowering agents were excluded to eliminate pharmacologic confounding; subsequently, independent associations of HDL-C and IR with depression were evaluated. Second, the interaction effects of HDL-C and IR on the odds of depression have been quantified. This investigation examines how IR—a core metabolic syndrome component—moderates HDL-C-depression association in a nationally representative sample.
Methods
Study design and participants
The NHANES is a nationally representative, cross-sectional surveillance system designed to assess population health metrics, nutritional status, and disease trends in the United States. This program employs a tripartite methodology: conducting interviews to capture sociodemographic, dietary, and health-related data; recording standardized physiological measurements during physical examinations; and performing biochemical assays to quantify serological and urinary biomarkers (including hematological parameters and metabolic profiles). For this investigation, data from seven consecutive NHANES cycles (2005–2018) were analyzed.
Main analysis variables
The HDL-C served as an exposure variable, and detailed analytical protocols for its quantification are comprehensively documented in the NHANES Laboratory Procedure Manual. Participants with a fasting duration of < 8.5 h (coded as fasting subsample weight = 0) were systematically excluded. Low HDL-C was defined as ≤ 1.0 mmol/L in males and ≤ 1.3 mmol/L in females [13, 14].
The outcome was depression, measured through the Patient Health Questionnaire-9 (PHQ-9), a validated 9-point instrument that quantifies symptom frequency (0 = never to 3 = nearly daily) during the preceding fortnight. Higher scores imply greater severity [15]. In this study, the participants were stratified into non-depression (score range: 0–9) and depression (score range: 10–27) groups based on a validated clinical threshold [9].
IR was quantified through the homeostasis model assessment 2 of insulin resistance (HOMA2-IR). This parameter was calculated from fasting glucose and insulin concentrations using the Oxford University HOMA Calculator [16]. Participants were classified as having IR when their HOMA2-IR measurements reached or exceeded the threshold of 2.5 [17].
Covariates
The covariates comprised age group, sex, race/ethnicity, education level, marital status, family poverty-to-income ratio (PIR) categories, body mass index (BMI) categories, smoking status, and alcohol consumption status [18]. Classification criteria are detailed in Table 1. Diabetes was defined as: (a) physician-diagnosed diabetes or current glucose-lowering medication use, or (b) laboratory-confirmed elevated metabolic parameters (fasting glucose ≥ 7.0 mmol/L or HbA1c ≥ 6.5%) [19]. Hypertension was defined as: physician diagnosis, current antihypertensive medication use, or blood pressure ≥ 130/80 mmHg [20]. Serum creatinine was used with the CKD-EPI 2021 equation to estimate glomerular filtration rate (eGFR) [21]. Antidepressant and antihyperlipidemic medication use was ascertained via NHANES Prescription Medications Questionnaire. Physical activity (PA) was quantified as weekly moderate-to-vigorous metabolic equivalent of task (MET-minutes) following WHO protocols. MET values were computed per established methods [22]. Activity levels were categorized according to American guidelines: inactive (0 MET-min/week), insufficiently active (1–599 MET-min/week), and sufficiently active (≥ 600 MET-min/week) [23]. Caffeine intake was assessed through dietary recall and categorized by daily intake: abstainers (0 mg), low (< 100 mg), moderate (100-<200 mg), and high (≥ 200 mg) [24].
Table 1.
Weighted distributions of characteristics across depression subgroups
| Variable | Overall | Non-depression | Depression | P |
|---|---|---|---|---|
| n | 50589415.92 | 48096424.71 | 2492991.21 | |
| Age (yeas) | 42.45 (15.47) | 42.51 (15.50) | 41.30 (14.89) | 0.161 |
| Age group (%) | 0.306 | |||
| ≤ 40 | 24972633.8 (49.4) | 23662384.1 (49.2) | 1310249.7 (52.6) | |
| > 40 | 25616782.1 (50.6) | 24434040.6 (50.8) | 1182741.5 (47.4) | |
| Sex (%) | < 0.001 | |||
| Female | 24450179.5 (48.3) | 22889911.1 (47.6) | 1560268.4 (62.6) | |
| Male | 26139236.5 (51.7) | 25206513.6 (52.4) | 932722.8 (37.4) | |
| Marital status (%) | < 0.001 | |||
| Never married | 10922323.7 (21.6) | 10289607.4 (21.4) | 632716.3 (25.4) | |
| Widowed or divorced or separated | 7254074.3 (14.3) | 6574089.0 (13.7) | 679985.3 (27.3) | |
| Married or living with partner | 32413017.9 (64.1) | 31232728.3 (64.9) | 1180289.6 (47.3) | |
| Race/Ethnicity (%) | < 0.001 | |||
| Hispanic | 7725084.0 (15.3) | 7279336.8 (15.1) | 445747.2 (17.9) | |
| Non-Hispanic White | 34028026.9 (67.3) | 32568169.5 (67.7) | 1459857.5 (58.6) | |
| Non-Hispanic Black | 5157519.1 (10.2) | 4745101.2 (9.9) | 412418.0 (16.5) | |
| Other Race | 3678785.9 (7.3) | 3503817.3 (7.3) | 174968.6 (7.0) | |
| Education level (%) | < 0.001 | |||
| Less than high school | 7012470.8 (13.9) | 6335639.4 (13.2) | 676831.4 (27.1) | |
| High school graduate | 11168177.2 (22.1) | 10459990.5 (21.7) | 708186.8 (28.4) | |
| Above high school | 32408767.9 (64.1) | 31300794.8 (65.1) | 1107973.1 (44.4) | |
| PIR Group (%) | < 0.001 | |||
| ≤ 3.0 | 25022216.5 (49.5) | 23047046.4 (47.9) | 1975170.1 (79.2) | |
| > 3.0 | 25567199.4 (50.5) | 25049378.3 (52.1) | 517821.1 (20.8) | |
| Smoking status (%) | < 0.001 | |||
| Never | 29320234.4 (58.0) | 28371777.0 (59.0) | 948457.4 (38.0) | |
| Former | 11159759.0 (22.1) | 10766842.1 (22.4) | 392916.9 (15.8) | |
| Current | 10109422.6 (20.0) | 8957805.6 (18.6) | 1151616.9 (46.2) | |
| Drinking status (%) | 0.135 | |||
| < 12 drinks/year | 18035158.0 (35.7) | 17057322.3 (35.5) | 977835.7 (39.2) | |
| ≥ 12 drinks/year | 32554257.9 (64.3) | 31039102.4 (64.5) | 1515155.5 (60.8) | |
| PA(MET-min/week) | < 0.001 | |||
| 0 | 8109500.6 (16.0) | 7435146.2 (15.5) | 674354.4 (27.1) | |
| 1–599 | 7269445.8 (14.4) | 6915449.3 (14.4) | 353996.6 (14.2) | |
| ≥ 600 | 35210469.5 (69.6) | 33745829.3 (70.2) | 1464640.2 (58.8) | |
| BMI (kg/m2) | 28.13 (6.56) | 28.10 (6.52) | 28.82 (7.19) | 0.078 |
| BMI group (%) | 0.025 | |||
| < 30 | 34835967.9 (68.9) | 33262308.9 (69.2) | 1573659.0 (63.1) | |
| ≥ 30 | 15753448.0 (31.1) | 14834115.8 (30.8) | 919332.2 (36.9) | |
| Caffeine intake (mg/d) | 165.64 (190.67) | 165.25 (189.68) | 173.32 (208.88) | 0.55 |
| Caffeine intake group (%) | 0.69 | |||
| No | 3610968.4 (7.1) | 3420230.1 (7.1) | 190738.3 (7.7) | |
| < 100 | 18836549.6 (37.2) | 17838689.7 (37.1) | 997859.9 (40.0) | |
| [100,200) | 12714180.9 (25.1) | 12101417.9 (25.2) | 612762.9 (24.6) | |
| ≥ 200 | 15427717.1 (30.5) | 14736086.9 (30.6) | 691630.1 (27.7) | |
| Hypertension (%) | 0.368 | |||
| No | 31366763.9 (62.0) | 29875917.6 (62.1) | 1490846.3 (59.8) | |
| Yes | 19222652.0 (38.0) | 18220507.1 (37.9) | 1002144.9 (40.2) | |
| HOMA2-IR | 1.25 (1.01) | 1.24 (1.01) | 1.40 (1.11) | 0.006 |
| IR (%) | 0.003 | |||
| No | 46037115.4 (91.0) | 43883873.6 (91.2) | 2153241.9 (86.4) | |
| Yes | 4552300.5 (9.0) | 4212551.1 (8.8) | 339749.3 (13.6) | |
| eGFR(mL/min/1.73 m²) | 100.87 (18.37) | 100.74 (18.41) | 103.34 (17.45) | 0.006 |
| eGFR Group (%) | 0.01 | |||
| < 60 | 1037621.8 (2.1) | 983681.0 (2.0) | 53940.9 (2.2) | |
| [60,89.9) | 12164955.0 (24.0) | 11726331.3 (24.4) | 438623.7 (17.6) | |
| ≥ 90 | 37386839.1 (73.9) | 35386412.5 (73.6) | 2000426.6 (80.2) | |
| HDL-C (mmol/L) | 1.42 (0.41) | 1.42 (0.41) | 1.39 (0.39) | 0.186 |
| HDL-C Group (%) | 0.006 | |||
| Low | 12311634.9 (24.3) | 11515248.0 (23.9) | 796387.0 (31.9) | |
| Normal | 38277781.0 (75.7) | 36581176.7 (76.1) | 1696604.3 (68.1) | |
| Glucose (mmol/L) | 5.45 (0.52) | 5.45 (0.52) | 5.44 (0.56) | 0.734 |
| Insulin(pmol/L) | 66.20 (55.77) | 65.79 (55.49) | 74.12 (60.32) | 0.006 |
| Survey Cycle (%) | 0.544 | |||
| 2005–2006 | 6238117.8 (12.3) | 6009395.4 (12.5) | 228722.4 (9.2) | |
| 2007–2008 | 7446565.2 (14.7) | 7090067.8 (14.7) | 356497.3 (14.3) | |
| 2009–2010 | 7330744.6 (14.5) | 6932054.5 (14.4) | 398690.1 (16.0) | |
| 2011–2012 | 7301276.0 (14.4) | 6884992.9 (14.3) | 416283.2 (16.7) | |
| 2013–2014 | 7347211.6 (14.5) | 7054687.9 (14.7) | 292523.8 (11.7) | |
| 2015–2016 | 7115982.2 (14.1) | 6745907.7 (14.0) | 370074.5 (14.8) | |
| 2017–2018 | 7809518.5 (15.4) | 7379318.6 (15.3) | 430199.9 (17.3) | |
All variables are displayed as means (standard deviations) or weighted frequencies and percentages (%)
Abbreviations: BMI Body mass index, eGFR Estimated Glomerular Filtration Rate, HDL-C High Density Lipoprotein Cholesterol, HOMA2-IR Homeostasis Model Assessment 2 of Insulin Resistance, IR Insulin Resistance, MET Metabolic Equivalent, PA Physical Activity, PIR Poverty-to-Income Ratio
Statistical analysis
The NHANES data analysis incorporated sampling weights from a 2-year Mobile Examination Center (MEC) subsample to maintain national representativeness [25]. Differences in variables were assessed using weighted linear regression or Rao-Scott adjusted chi-square tests.
Three sequential survey-weighted generalized linear models (GLMs) were constructed to evaluate associations of HDL-C or IR with depression: Model 1 (unadjusted); Model 2 (sociodemographic-adjusted); and Model 3 (comprehensive; Model 2 covariates plus smoking/drinking status, caffeine intake, PA, eGFR, and hypertension). Model 3 employed survey-weighted restricted cubic spline (RCS) regression with pre-smoothing to characterize associations. Overall and non-linear P-values were calculated for trend significance testing.
Building upon Model 3, survey-weighted GLMs were performed using the survey R package, with depression status (present vs. absent) as the response variable and incorporating interaction terms between HDL-C (low vs. normal) and IR (yes vs. no). The interactionR package was used to assess multiplicative and additive interactions. A P < 0.05 on the multiplicative scale indicated significant multiplicative interaction. Additive interaction significance required P < 0.05 for either the relative excess risk due to interaction (RERI) or the attributable proportion due to interaction (AP). Finally, subgroup analyses stratified by IR status (yes vs. no) were performed in the Model 3 to evaluate HDL-C and depression associations. Interaction effects were formally tested using likelihood ratio tests.
A series of sensitivity analyses were performed: First, IR status was redefined using alternative thresholds (HOMA2-IR ≥ 2.0, ≥ 2.3, and ≥ 2.7) to evaluate the effect of diagnostic criteria [26]. Second, total cholesterol (TC) and triglyceride (TG) were incorporated into the base model (Model 3) to account for residual confounding. Third, analyses were performed using 2-year MEC weights based on the fasting subsample, supplemented by unweighted analyses to assess sensitivity to the sampling design. Finally, extended analyses were conducted by separately including individuals with diabetes or those taking antihyperlipidemic medications, with diabetes status (yes/no) and antihyperlipidemic drug use (yes/no) added as additional covariates in Model 3. All analyses were performed using R Statistical Software (version 4.3.3) and the Free Statistics platform (version 2.3).
Results
Baseline characteristics
Following Fig. 1 screening protocols, 7,779 eligible participants formed the analytical cohort. Table 1 presents the weighted baseline characteristics, showing significant intergroup differences between depression and non-depression participants. The depression group revealed significant sex disparity, with 62.6% female (P < 0.001). Age distribution did not differ significantly (P = 0.306), whereas significant differences existed in marital status, race, education, PIR, smoking status, and MET scores (all P < 0.05). Significant variations also occurred in the BMI categories, eGFR, HOMA2-IR, and HDL-C groups (P < 0.05), indicating the multifactorial associations of sociodemographic, behavioral, and cardiometabolic factors with depression.
Fig. 1.
Participant inclusion flowchart.Abbreviations: BMI: Body mass index; eGFR: Estimated Glomerular Filtration Rate; HDL-C: High Density Lipoprotein Cholesterol; NHANES: National Health and Nutrition Examination Survey; PA, Physical Activity; PHQ-9: Patient Health Questionnaire 9; PIR: Poverty-to-Income Ratio
The respective associations of HDL-C and IR with depression
Survey-weighted GLM assessed associations of HDL-C or IR with depression (Table 2). In unadjusted analysis, normal HDL-C was negatively associated with depression (odds ratio [OR] = 0.67, 95% confidence interval [CI]: 0.51–0.89, P = 0.006), while continuous HDL-C showed no association (P = 0.202). Per 1-unit HOMA2-IR increase conferred 13% higher depression odds (OR = 1.13, 95% CI: 1.05–1.22; P = 0.001); the IR group had 64% increased odds (OR = 1.64, 95% CI: 1.18–2.29; P = 0.003). However, these associations became non-significant after covariate adjustment. Subsequent survey-weighted RCS analyses in Model 3 revealed no associations between HDL-C or IR and depression (Fig. 2; all P for non-linearity > 0.05; all P for overall > 0.05).
Table 2.
The associations of HDL-C or IR with depression
| Variables | Model 1a | Model 2b | Model 3c | |||
|---|---|---|---|---|---|---|
| OR (95%CI) d | P | OR (95%CI) | P | OR (95%CI) | P | |
| HDL-C (mmol/L) | 0.82 (0.61, 1.11) | 0.202 | 0.78 (0.53, 1.14) | 0.192 | 0.92 (0.66, 1.27) | 0.611 |
| HDL-C (group) | ||||||
| Low | Ref | Ref | Ref | |||
| Normal | 0.67 (0.51, 0.89) | 0.006 | 0.88 (0.66,1.18) | 0.389 | 1.04 (0.78, 1.40) | 0.77 |
| HOMA2-IR | 1.13 (1.05, 1.22) | 0.001 | 1.06 (0.97, 1.17) | 0.178 | 1.07 (0.97, 1.17) | 0.174 |
| IR (group) | ||||||
| No | Ref | Ref | Ref | |||
| Yes | 1.64 (1.18, 2.29) | 0.003 | 1.38 (0.94, 2.01) | 0.097 | 1.34 (0.91, 1.98) | 0.136 |
a. Unadjusted
b. Adjusted for sex, age categories, race/ethnicity, education level, marital status, PIR groups, and BMI categories;
c. Model 2 + smoking and drinking status, caffeine consumption, PA, eGFR, and hypertension
d. OR and 95% CI represent the odds of depression per 1-unit increase in HDL-C or HOMA2-IR (continuous variables) or relative to the reference group
Abbreviations: CI Confidence Intervals, HDL-C High Density Lipoprotein Cholesterol, HOMA2-IR Homeostasis Model Assessment 2 of Insulin Resistance, IR Insulin Resistance, OR Odds Ratios, Ref Reference group
Fig. 2.
Association Between HDL-C or HOMA2-IR and odds of depression. Panels: (a) HDL-C; (b) HOMA2-IR.Abbreviations: CI, Confidence Intervals; HDL-C, High Density Lipoprotein Cholesterol; HOMA2-IR, Homeostasis Model Assessment 2 of Insulin Resistance; OR, Odds ratios
Interaction effects of HDL-C and IR on the odds of depression
Significant HDL-C × IR interactions on depression occurred in fully adjusted models (Model 3, Table 3). Multiplicative (OR = 3.02, 95% CI: 1.43–6.36; P = 0.004) and additive interactions were significant: RERI = 1.40 (95% CI: 0.33–2.47; P = 0.005) and AP = 0.68 (95% CI: 0.40–0.96; P < 0.001). Subgroup analyses demonstrated significant effect modification by IR status. For continuous HDL-C, associations with depression significantly differed by IR status (P interaction = 0.024): There was no association in non-IR individuals (OR = 0.86, 95% CI: 0.60–1.23; P = 0.40) but a positive association in IR individuals (OR = 4.66, 95% CI: 1.23–17.59; P = 0.02). For categorical HDL-C (normal vs. low), a significant interaction existed (P interaction = 0.005): normal HDL-C showed a null association without IR (OR = 0.91, 95% CI: 0.65–1.28; P = 0.58) but significantly elevated odds with IR (OR = 3.79, 95% CI: 1.79–8.03; P < 0.001).
Table 3.
Interaction effects of HDL-C and IR on depression
| Variable | Estimates | 95%CI | P | P for interaction |
|---|---|---|---|---|
| Multiplicative scale | 3.02 | (1.43,6.36) | 0.004 | |
| RERI | 1.40 | (0.33,2.47) | 0.005 | |
| AP | 0.68 | (0.40,0.96) | < 0.001 | |
| HDL-C on Depressiona | 0.024 | |||
| IR (No) | 0.86 | (0.60, 1.23) | 0.4 | |
| IR (Yes) | 4.66 | (1.23, 17.59) | 0.02 | |
| HDL-C on Depressionb | 0.005 | |||
| IR (No) | ||||
| Low HDL-C | Ref | |||
| Normal HDL-C | 0.91 | (0.65, 1.28) | 0.58 | |
| IR (Yes) | ||||
| Low HDL-C | Ref | |||
| Normal HDL-C | 3.79 | (1.79, 8.03) | < 0.001 |
a: Association between continuous HDL-C (per 1-unit increase) and depression
b: Association between categorical HDL-C (Normal vs. Low) and depression
All analyses are based on the Model 3
Abbreviations: AP Attributable Proportion due to interaction, CI Confidence Intervals, HDL-C High Density Lipoprotein Cholesterol, IR Insulin Resistance, Ref Reference group, RERI Relative Excess Risk due to Interaction
Sensitivity analysis
Across all tested scenarios—including varying HOMA2-IR thresholds (≥ 2.0, ≥ 2.3, or ≥ 2.7), additional adjustment for blood lipids, alternative weighting methods, and the inclusion of special populations—the multiplicative interaction terms remained statistically significant (range: 1.91–3.00, all P < 0.05). Synergistic effects were further confirmed by positive RERI (0.70–1.43, P ≤ 0.02) and AP (0.49–0.67, P ≤ 0.002) in additive interactions (Table 4). These analyses confirmed the robustness of the HDL-C × IR synergistic effect on depression observed in the initial analyses.
Table 4.
Interaction between HDL-C and IR on depression under different condition
| Condition | Multiplicative scale | RERI | AP | |||
|---|---|---|---|---|---|---|
| Estimates (95%CI) |
P | Estimates (95%CI) |
P | Estimates (95%CI) |
P | |
|
IR was defined as HOMA2-IR ≥ 2.0 |
2.18 (1.05, 4.51) | 0.039 | 0.89 (0.09, 1.69) | 0.014 | 0.55 (0.17, 0.93) | 0.002 |
|
IR was defined as HOMA2-IR ≥ 2.3 |
2.67 (1.22, 5.88) | 0.016 | 1.17 (0.18, 2.15) | 0.01 | 0.64 (0.31, 0.97) | < 0.001 |
|
IR was defined as HOMA2-IR ≥ 2.7 |
2.40 (1.11, 5.17) | 0.028 | 1.15 (0.09, 2.19) | 0.016 | 0.59 (0.25, 0.93) | < 0.001 |
| Further adjustment of TC and TG based on Model 3 | 3.00 (1.42, 6.33) | 0.005 | 1.43 (0.35, 2.52) | 0.004 | 0.67 (0.39, 0.95) | < 0.001 |
| Calculated based on Fasting Subsample 2 Year MEC weight | 2.77 (1.34, 5.72) | 0.007 | 1.26 (0.28, 2.25) | 0.006 | 0.65 (0.35, 0.95) | < 0.001 |
| Unweighted | 1.91 (1.07, 3.43) | 0.029 | 0.70 (0.04, 1.35) | 0.018 | 0.49 (0.15, 0.83) | 0.002 |
| Inclusion of individuals with diabetes (n = 8776) | 2.06 (1.07, 3.98) | 0.033 | 0.89 (0.04, 1.74) | 0.02 | 0.52 (0.17, 0.86) | 0.002 |
| Inclusion of individuals taking antihyperlipidemic drugs (n = 8910) | 2.48 (1.21, 5.10) | 0.015 | 1.09 (0.16, 1.99) | 0.009 | 0.61 (0.29, 0.93) | < 0.001 |
Abbreviations: AP Attributable Proportion due to interaction, CI Confidence Intervals, IR Insulin Resistance, MEC Mobile Examination Center, RERI Relative Excess Risk due to Interaction, TC Total Cholesterol, TG Triglycerides
Discussion
Individual analyses revealed no significant association between HDL-C levels (categorical or continuous) and depression. However, HDL-C significantly increased the odds of depression when IR was present. Given variability in HOMA2-IR thresholds [26], the sensitivity analyses defined IR using cutoffs of 2.0, 2.3, and 2.7. HDL-C × IR interactions remained consistently significant and synergistic across thresholds. The inclusion of participants with lipid-lowering agents (which may alter HDL-C) or with diabetes (which may alter IR) did not change the direction of the interactions. Collectively, these data suggest that elevated HDL-C may confer increased odds of depression in the context of coexisting IR.
The dynamic relationship between HDL-C and depression in the context of IR may be explained by the following mechanisms. Under physiological conditions, cholesterol derived from HDL-C serves as a substrate for basal cortisol synthesis in the adrenal glands [27]. Population-based studies have demonstrated a significant, positive correlation between plasma cortisol and HDL-C levels [28, 29]. Dysregulation of cortisol is a contributor to depression [30]. HDL particles exhibit potent anti-inflammatory properties [31, 32], while systemic inflammation is a hallmark feature of depression [33]. Under physiological conditions, the HDL-C concentration may reflect HDL’s anti-inflammatory effects, mediated by the reverse cholesterol transport function [34]. Thus, the potential pro-depressive effects of HDL-C (through its role in cortisol synthesis) and its antidepressant effects (through anti-inflammatory mechanisms) may counteract one another. This balance may explain the lack of a significant association between HDL-C and depression in individuals without IR.
IR causes significant structural and functional alterations in HDL particles [35]. Animal studies show that intestinal-derived HDL exhibits reduced apolipoprotein A-I (apoA-I) content during IR [36]. This finding has been corroborated in humans: HOMA2-IR positively correlates with the apoB/apoA-I ratio [37, 38]. HDL’s anti-inflammatory and antioxidant capacities are primarily mediated by paraoxonase 1 (PON1) [39]. PON1 activity decreases during IR [40, 41]. Depressed patients also exhibited significantly lowered plasma PON1 activity compared to healthy controls [42, 43]. Moreover, dysfunctional HDL in obesity paradoxically acquires pro-inflammatory activity [44]. Consequently, IR impairs HDL’s anti-inflammatory capacity. This dysfunction decouples HDL-C concentration from its anti-inflammatory activity, while HDL-C remains a cortisol synthesis substrate. Thus, elevated HDL-C during IR may increase the odds of depression through this mechanism.
Strengths and limitations
This study has several notable strengths. First, it utilized a large, nationally representative sample from the NHANES (2005–2018), applying rigorous exclusion criteria, comprehensive adjustments for potential confounders, and extensive sensitivity analyses. These approaches collectively strengthen the generalizability and robustness of the findings to the U.S. population. Second, the results challenge the conventional perception of HDL-C as universally cardioprotective “good cholesterol,” revealing its context-dependent association with the odds of neuropsychiatric disorders. Finally, the findings underscore that HDL-C cannot simply be used as a substitute for HDL function in clinical practice, especially in the presence of IR. Indicators that can more comprehensively and stably reflect the function of HDL need to be further explored.
The study also has several limitations. The cross-sectional design precludes definitive conclusions regarding causality between HDL-C and depression, underscoring the need for validation in prospective cohort studies. Despite the validated use of the PHQ-9 in population surveys, reliance on self-reported scores may introduce the potential for measurement bias. Moreover, single-timepoint measurements of HDL-C may not adequately reflect chronic metabolic status. Future research should address these limitations to strengthen the clinical and mechanistic interpretations of the findings.
Conclusion
Elevated HDL-C levels, when accompanied by IR, are significantly associated with a greater likelihood of depression. The finding highlights the context-dependent link between HDL-C and the odds of depression, which specifically depends on IR status. Therefore, clinicians should prioritize depression screening in individuals with both high HDL-C levels and IR or related conditions. Large-scale prospective studies are needed to establish causality. Such evidence could reshape depression prevention strategies by enabling more targeted interventions.
Acknowledgements
The authors express profound gratitude to the National Center for Health Statistics research team for their rigorous execution of complex survey protocols and commitment to open data dissemination through the NHANES repository.
Abbreviations
- AP
Attributable Proportion due to interaction
- BMI
Body Mass Index
- CI
Confidence Interval
- CKD-EPI
Chronic Kidney Disease Epidemiology Collaboration
- eGFR
Estimated Glomerular Filtration Rate
- GLM
Generalized Linear Models
- HDL-C
High Density Lipoprotein Cholesterol
- HOMA2-IR
Homeostasis Model Assessment 2 of Insulin Resistance
- IR
Insulin Resistance
- MEC
Mobile Examination Center
- MET
Metabolic Equivalent
- NHANES
National Health and Nutrition Examination Survey
- OR
Odds Ratio
- PA
Physical Activity
- PHQ-9
Patient Health Questionnaire 9
- PIR
Poverty-to-Income Ratio
- RCS
Restricted Cubic Spline
- RERI
Relative Excess Risk due to Interaction
- TC
Total Cholesterol
- TG
Triglycerides
Authors’ contributions
Liang Rao: Writing – original draft & Data curation; Jing Li: Formal Analysis & Data curation; Xiaohua Xian: Writing – review & editing; Yanyi Hu: Validation & Funding acquisition; Yin Xian: Conceptualization & Data curation & Formal Analysis & Funding acquisition &Visualization.
Funding
This work was supported by Bureau of Science and Technology Nanchong City [grant numbers 22YYJCYJ0054, 23YYJCYJ0151], Natural Science Foundation of Sichuan Province [grant numbers 2023JDRC0091], the Civil Affairs Fund of the Sichuan Government (China), and the Minors Protection Care Foundation of the Sichuan Charity Federation (China).
Data availability
The analytical data for this investigation were obtained from the publicly accessible data repository of (https://wwwn.cdc.gov/Nchs/Nhanes/).
Declarations
Ethics approval and consent to participate
The National Center for Health Statistics Ethics Review Board approved the survey, and all participants provided written informed consent. This research was conducted in accordance with the Declaration of Helsinki.
Consent for publication
Not applicable.
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.
Liang Rao and Jing Li are co-first authors.
Contributor Information
Yanyi Hu, Email: tnb8076816@163.com.
Yin Xian, Email: xianyin@ncsxyy.com.
References
- 1.Shi J, Han X, Liao Y, Zhao H, Fan B, Zhang H, et al. Associations of stressful life events with subthreshold depressive symptoms and major depressive disorder: the moderating role of gender. J Affect Disord. 2023;325:588–95. [DOI] [PubMed] [Google Scholar]
- 2.Global regional. and national burden of 12 mental disorders in 204 countries and territories, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet Psychiatry 2022; 9:137–150. [DOI] [PMC free article] [PubMed]
- 3.Anand A, Mathew SJ, Sanacora G, Murrough JW, Goes FS, Altinay M, et al. Ketamine versus ECT for nonpsychotic treatment-resistant major depression. N Engl J Med. 2023;388:2315–25. [DOI] [PubMed] [Google Scholar]
- 4.Sirtori CR, Corsini A, Ruscica M. The role of high-density lipoprotein cholesterol in 2022. Curr Atheroscler Rep. 2022;24:365–77. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Groenen AG, Halmos B, Tall AR, Westerterp M. Cholesterol efflux pathways, inflammation, and atherosclerosis. Crit Rev Biochem Mol Biol. 2021;56:426–39. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Beurel E, Toups M, Nemeroff CB. The bidirectional relationship of depression and inflammation: double trouble. Neuron. 2020;107:234–56. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Oh J, Kim TS. Serum lipid levels in depression and suicidality: the Korea National health and nutrition examination survey (KNHANES) 2014. J Affect Disord. 2017;213:51–8. [DOI] [PubMed] [Google Scholar]
- 8.Zhang Q, Liu Z, Wang Q, Li X. Low cholesterol is not associated with depression: data from the 2005–2018 National health and nutrition examination survey. Lipids Health Dis. 2022;21:35. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Zhong X, Ming J, Li C. Association between dyslipidemia and depression: a cross-sectional analysis of NHANES data from 2007 to 2018. BMC Psychiatry. 2024;24:893. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Ertek S. High-density lipoprotein (HDL) dysfunction and the future of HDL. Curr Vasc Pharmacol. 2018;16:490–8. [DOI] [PubMed] [Google Scholar]
- 11.Piko P, Jenei T, Kosa Z, Sandor J, Kovacs N, Seres I, Paragh G, Adany R. Association of HDL subfraction profile with the progression of insulin resistance. Int J Mol Sci. 2023;24(17):13563. [DOI] [PMC free article] [PubMed]
- 12.Chapman MJ. HDL functionality in type 1 and type 2 diabetes: new insights. Curr Opin Endocrinol Diabetes Obes. 2022;29:112–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Lam DW, LeRoith D et al. Metabolic Syndrome. In: Feingold KR, Ahmed SF, Anawalt B, eds. Endotext. South Dartmouth (MA): MDText.com, Inc.; February 11, 2019.
- 14.Jin ES, Shim JS, Kim SE, Bae JH, Kang S, Won JC, et al. Dyslipidemia fact sheet in South Korea, 2022. Diabetes Metab J. 2023;47:632–42. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Zimmerman M. Using the 9-Item patient health questionnaire to screen for and monitor depression. JAMA. 2019;322:2125–6. [DOI] [PubMed] [Google Scholar]
- 16.Li X, Wang X, Park SK. Associations between rice consumption, arsenic metabolism, and insulin resistance in adults without diabetes. Int J Hyg Environ Health. 2021;237:113834. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Friedrich N, Thuesen B, Jørgensen T, Juul A, Spielhagen C, Wallaschofksi H, et al. The association between IGF-I and insulin resistance: a general population study in Danish adults. Diabetes Care. 2012;35:768–73. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Shen Y, Zhu Z, Bi X, Shen Y, Shen A, Deng B, et al. Association between insulin resistance indices and kidney stones: results from the 2015–2018 National health and nutrition examination survey. Front Nutr. 2024;11:1444049. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Kianersi S, Liu Y, Guasch-Ferré M. Chronotype, unhealthy lifestyle, and diabetes risk in middle-aged U.S. women: a prospective cohort study. Ann Intern Med. 2023;176:1330–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Huang Z. Association between blood lead level with high blood pressure in US (NHANES 1999–2018). Front Public Health. 2022;10:836357. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Grams ME, Brunskill NJ, Ballew SH, Sang Y, Coresh J, Matsushita K, et al. The kidney failure risk equation: evaluation of novel input variables including eGFR estimated using the CKD-EPI 2021 equation in 59 cohorts. J Am Soc Nephrol. 2023;34:482–94. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Wei X, Min Y, Xiang Z, Zeng Y, Wang J, Liu L. Joint association of physical activity and dietary quality with survival among US cancer survivors: a population-based cohort study. Int J Surg. 2024;110:5585–94. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Piercy KL, Troiano RP, Ballard RM, Carlson SA, Fulton JE, Galuska DA, et al. The physical activity guidelines for Americans. JAMA. 2018;320:2020–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Lin P, Liang Z, Wang M. Caffeine consumption and mortality in populations with different weight statuses: an analysis of NHANES 1999–2014. Nutrition. 2022;102:111731. [DOI] [PubMed] [Google Scholar]
- 25.Johnson CL, Paulose-Ram R, Ogden CL, Carroll MD, Kruszon-Moran D, Dohrmann SM, et al. National health and nutrition examination survey: analytic guidelines, 1999–2010. Vital Health Stat. 2013;2:1–24. [PubMed] [Google Scholar]
- 26.Tahapary DL, Pratisthita LB, Fitri NA, Marcella C, Wafa S, Kurniawan F, et al. Challenges in the diagnosis of insulin resistance: focusing on the role of HOMA-IR and tryglyceride/glucose index. Diabetes Metab Syndr. 2022;16:102581. [DOI] [PubMed] [Google Scholar]
- 27.Bochem AE, Holleboom AG, Romijn JA, Hoekstra M, Dallinga-Thie GM, Motazacker MM, et al. High density lipoprotein as a source of cholesterol for adrenal steroidogenesis: a study in individuals with low plasma HDL-C. J Lipid Res. 2013;54:1698–704. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Le-Ha C, Herbison CE, Beilin LJ, Burrows S, Henley DE, Lye SJ, et al. Hypothalamic-pituitary-adrenal axis activity under resting conditions and cardiovascular risk factors in adolescents. Psychoneuroendocrinology. 2016;66:118–24. [DOI] [PubMed] [Google Scholar]
- 29.Hayashi R, Tamada D, Murata M, Kitamura T, Mukai K, Maeda N, et al. Glucocorticoid replacement affects serum adiponectin levels and HDL-C in patients with secondary adrenal insufficiency. J Clin Endocrinol Metab. 2019;104:5814–22. [DOI] [PubMed] [Google Scholar]
- 30.Zajkowska Z, Gullett N, Walsh A, Zonca V, Pedersen GA, Souza L, et al. Cortisol and development of depression in adolescence and young adulthood - a systematic review and meta-analysis. Psychoneuroendocrinology. 2022;136:105625. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Yu M, Dorsey KH, Halseth T, Schwendeman A. Enhancement of anti-inflammatory effects of synthetic high-density lipoproteins by incorporation of anionic lipids. Mol Pharm. 2023;20:5454–62. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Yang L, Wang Y, Xu Y, Li K, Yin R, Zhang L, et al. Angptl3 is a novel HDL component that regulates HDL function. J Transl Med. 2024;22:263. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Meng X, Han L, Fu J, Hu C, Lu Y. Associations between metabolic syndrome and depression, and the mediating role of inflammation: based on the NHANES database. J Affect Disord. 2025;375:214–21. [DOI] [PubMed] [Google Scholar]
- 34.Kang H, Song J, Cheng Y. HDL regulates the risk of cardiometabolic and inflammatory-related diseases: focusing on cholesterol efflux capacity. Int Immunopharmacol. 2024;138:112622. [DOI] [PubMed] [Google Scholar]
- 35.Lui DTW, Tan KCB. High-density lipoprotein in diabetes: structural and functional relevance. J Diabetes Investig. 2024;15:805–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Mangat R, Borthwick F, Haase T, Jacome M, Nelson R, Kontush A, et al. Intestinal lymphatic HDL miR-223 and ApoA-I are reduced during insulin resistance and restored with niacin. FASEB J. 2018;32:1602–12. [DOI] [PubMed] [Google Scholar]
- 37.Ying X, Qian Y, Jiang Y, Jiang Z, Song Z, Zhao C. Association of the apolipoprotein B/apolipoprotein A-I ratio and low-density lipoprotein cholesterol with insulin resistance in a Chinese population with abdominal obesity. Acta Diabetol. 2012;49:465–72. [DOI] [PubMed] [Google Scholar]
- 38.Makaridze Z, Giorgadze E, Asatiani K. Association of the apolipoprotein b/apolipoprotein a-I ratio, metabolic syndrome components, total cholesterol, and low-density lipoprotein cholesterol with insulin resistance in the population of Georgia. Int J Endocrinol. 2014;2014:925650. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Jakubowski H. The molecular bases of Anti-Oxidative and Anti-Inflammatory properties of paraoxonase 1. Antioxid (Basel). 2024;13(11):1292. [DOI] [PMC free article] [PubMed]
- 40.Doğan K, Şeneş M, Karaca A, Kayalp D, Kan S, Gülçelik NE, et al. HDL subgroups and their paraoxonase-1 activity in the obese, overweight and normal weight subjects. Int J Clin Pract. 2021;75:e14969. [DOI] [PubMed] [Google Scholar]
- 41.Krzystek-Korpacka M, Patryn E, Hotowy K, Czapińska E, Majda J, Kustrzeba-Wójcicka I, Noczyńska A, Gamian A. Paraoxonase (PON)-1 activity in overweight and obese children and adolescents: association with obesity-related inflammation and oxidative stress. Adv Clin Exp Med. 2013;22:229–36. [PubMed] [Google Scholar]
- 42.Bortolasci CC, Vargas HO, Souza-Nogueira A, Barbosa DS, Moreira EG, Nunes SO, Berk M, Dodd S, Maes M. Lowered plasma paraoxonase (PON)1 activity is a trait marker of major depression and PON1 Q192R gene polymorphism-smoking interactions differentially predict the odds of major depression and bipolar disorder. J Affect Disord. 2014;159:23–30. [DOI] [PubMed] [Google Scholar]
- 43.Moreira EG, Correia DG, Bonifácio KL, Moraes JB, Cavicchioli FL, Nunes CS, et al. Lowered PON1 activities are strongly associated with depression and bipolar disorder, recurrence of (hypo)mania and depression, increased disability and lowered quality of life. World J Biol Psychiatry. 2019;20:368–80. [DOI] [PubMed] [Google Scholar]
- 44.Roberts CK, Ng C, Hama S, Eliseo AJ, Barnard RJ. Effect of a short-term diet and exercise intervention on inflammatory/anti-inflammatory properties of HDL in overweight/obese men with cardiovascular risk factors. J Appl Physiol (1985). 2006;101:1727–32. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The analytical data for this investigation were obtained from the publicly accessible data repository of (https://wwwn.cdc.gov/Nchs/Nhanes/).