- Top of page
Aims: The purpose of the present study was to examine whether the association between depression and the serum high-density lipoprotein cholesterol (HDL-C) is modified by symptom duration.
Methods: Depressed patients (n = 88) and an age- and sex-matched group of healthy general population controls (n = 88) underwent a Structured Clinical Interview for DSM-IV (SCID), and depressed participants reported the duration of their symptoms. The serum levels of total cholesterol (TC), HDL-C, low-density lipoprotein cholesterol (LDL-C), triglycerides (TG) and non-HDL, and the ratios of LDL-C/HDL and TC/HDL-C were assessed.
Results: Major depressive disorder (MDD) subjects with a long symptom duration (≥3 years) had lower levels of HDL-C compared with healthy controls or MDD subjects with a symptom duration <3 years. The likelihood for long symptom duration doubled for each 0.5-mmol/L decrease in HDL-C levels in regression models adjusted for age, gender, marital status, overweight, symptom severity, alcohol consumption, smoking, physical exercise, medication use, and non-HDL-C (P < 0.05).
Conclusions: These findings suggest that a low serum HDL-C level, a risk factor for coronary heart disease, is specifically associated with long-term depressive symptomatology.
HIGH-DENSITY LIPOPROTEIN (HDL) is one of the carrier proteins responsible for the transportation of serum lipids, and high serum levels of this are associated with a decreased risk of vascular illnesses.1 It not only locally calms inflammation in the vascular epithelium, and is thus able to regress the formation of atherosclerotic plaques, but it also has systemic anti-inflammatory effects.2 In contrast, pronounced low-grade inflammation has been demonstrated to be associated with depression.3 Therefore, the role of HDL cholesterol (HDL-C) alterations in depression is of increasing interest in the light of the newer pathophysiological theories of depression.
Previous investigations into depression-related HDL-C alterations have presented discrepant findings.4–6 Both low4,6 and high5 levels of HDL-C have been associated with depression. Nevertheless, recent data suggest that there are factors that could explain these discrepant observations. Our group has reported lowered levels of HDL-C in general population subjects with chronic depression.7 Moreover, the duration of depressive symptoms has previously been shown to modify biological correlates in depressive disorders.8 We hypothesized that symptom duration could also modify the association between major depressive disorder (MDD) and HDL-C, and thus we examined lipid levels in MDD patients, divided into two symptom duration groups, and in healthy controls.
- Top of page
The present investigation focused on psychiatric patients who were referred to a psychiatric examination ward for diagnostic re-evaluation and treatment plan formulation. All patients entering the ward between September 2004 and September 2005 were offered the possibility to participate in the study, and a total of 150 patients were enrolled. The Structured Clinical Interview for DSM-IV (SCID-I)9 was used for all recruited patients and were applied by trained, experienced staff interviewers. A total of 105 patients were diagnosed to have MDD. Of these, 17 were excluded due to incomplete data, and thus the final sample consisted of 88 patients. Depressive symptoms were evaluated with the Beck Depression Inventory (BDI)10 at the beginning of hospitalization, simultaneously with the collection of the background data. Sociodemographic data, duration of symptoms, lifestyle factors (alcohol consumption, smoking habits and the frequency of physical exercise), and the use of antidepressant, antipsychotic and lipid-lowering medications were also investigated with a questionnaire. The MDD patients reported their symptom duration as a categorical variable (<6 months/6–<12 months/12 months–2 years/>2 years), and those reporting a symptom duration of ≥2 years were also requested to provide a numerical duration estimate. The 3-year cut-off was formed by combining the data from both of the aforedescribed questions; the subjects who reported that they had (i) a symptom duration >2 years; and (ii) gave a numerical duration estimate of ≥3 years were included in the long symptom duration (≥3 years) group.
Age- and sex-matched healthy controls (n = 88) were randomly selected from a general population database that was collected in the same area as the patient data, and has been thoroughly described earlier.7,11 Briefly, the database was collected as a part of a four-phase general population study, with baseline data collection in 1998 and follow ups in 1999 and 2001. The fourth phase of the study was conducted between September 2004 and December 2005, that is, at the same time as the collection of the patient data. All controls reported low BDI scores (i.e. BDI < 10) in all previous follow ups, underwent a SCID-I interview to confirm their euthymic state, and also completed a questionnaire reporting data on sociodemographic and lifestyle factors.
Approval for both studies was obtained from the Ethics Committee of Kuopio University Hospital and the University of Kuopio, and both study protocols were in accordance with the latest version of the Declaration of Helsinki. All participants provided written informed consent before entering the studies.
Height and bodyweight were measured in light clothing without shoes, and the body mass index (BMI; kg/m2) was calculated. All participants were requested to fast for 12 h before sampling, and venous blood samples were drawn at 08.00 hours. Serum total cholesterol (TC; reference value < 5.0 mmol/L), HDL-C (reference value > 1.0 mmol/L) and triglyceride (TG; reference value < 2.0 mmol/L) measurements were carried out according to the routine protocol in the accredited Kuopio University Hospital medical laboratory. Enzymatic methods (Thermo Electron, Vantaa, Finland) were used for all measurements (TC: Konelab Cholesterol, code 981812; HDL-C: Konelab HDL-Cholesterol, code 981655; TG: Konelab Triglycerides, code 981301). The total variations of the utilized methods were 1.6%, 3.7% and 4.8%, respectively. The samples were analyzed using a Konelab 60i Clinical Chemistry Analyzer (Thermo Electron). The level of low-density lipoprotein cholesterol (LDL-C; reference value < 3.0 mmol/L) was calculated according to the Friedewald formula.12 In addition, we calculated LDL-C/HDL-C and TC/HDL-C ratios, and non-HDL-C (TC minus HDL-C). Non-HDL-C enables assessment of the impact of both LDL-C and TG in multivariate models, because it reflects the LDL-C level and is also a measure of TG catabolism.
In order to test the association between symptom duration and the HDL-C levels, we used a duration variable with a cut-off of 3 years (long duration: ≥3 years; short duration: <3 years). The cut-off was based on statistical optimization, providing as equal a division of the MDD study subjects into symptom duration groups as possible, and thus also maximizing the statistical power in the analyses. The analyses were performed (i) for the whole sample (the two MDD groups and healthy controls); and then (ii) further for the two MDD groups. First, differences between the three groups were assessed using the χ2 test or Fisher's exact test for categorical variables, and the Kruskal–Wallis test and analysis of variance (anova) for continuous variables. Parametric tests were used when the distribution of variables was normal and non-parametric tests when the distribution was skewed. Second, the likelihood for belonging to the long symptom duration group was examined with several relevant covariates (age, gender, marital status, overweight, symptom severity, non-HDL-C, physical exercise, alcohol consumption, smoking, and the use of antidepressant, antipsychotic and lipid-lowering medications) on logistic regression (method: enter). The basic model (model 1) was adjusted for age, gender, marital status, overweight, symptom severity, and non-HDL. In model 2, model 1 was further adjusted for physical exercise (i.e. model 1 covariates and physical exercise). In model 3, model 1 was further adjusted for medication use (i.e. model 1 covariates and the use of antidepressant, antipsychotic and lipid-lowering medications), and in model 4, model 1 was further adjusted for alcohol consumption and smoking (i.e. model 1 covariates and alcohol consumption and smoking). The same regression analyses were also performed for other lipids and lipid indices that were observed to differ among the examined groups. In these models, HDL-C and non-HDL-C were omitted due to physiological redundancy. Two-tailed P < 0.05 was considered to indicate statistical significance in all analyses.
- Top of page
The healthy controls were more often married or living with a partner than the MDD subjects, who in turn were more frequently smokers, suffered from overweight, and had higher BDI scores. Among the three study groups, the MDD subjects with long symptom duration had the lowest HDL-C levels on anova adjusted for age, gender, and non-HDL-C (Table 1). When specifically comparing the two MDD groups, each 0.5-mmol/L decrease in the level of HDL-C doubled the likelihood for belonging to the long symptom duration group in a model adjusted for age, gender, marital status, overweight, symptom severity, and non-HDL (model 1, odds ratio [OR], 2.20; P = 0.016; Table 2). These findings remained unaltered in all further adjusted models (model 2, OR, 2.12; P = 0.022; model 3, OR, 2.14; P = 0.025; model 4, OR 2.09, P = 0.047). Each 0.5-step increase in TC/HDL-C or LDL-C/HDL-C was also independently associated with belonging to the long symptom duration group in model 1 (OR, 1.20; P = 0.040; OR, 1.16; P = 0.044, respectively; Table 2). These findings remained significant in model 2 (TC/HDL-C, OR, 1.21; P = 0.037; LDL-C/HDL-C, OR, 1.16; P = 0.040), but not in models 3 (TC/HDL-C, OR, 1.19; P = 0.052; LDL-C/HDL-C, OR, 1.15; P = 0.055) or 4 (TC/HDL-C, OR, 1.20; P = 0.059; LDL-C/HDL-C, OR, 1.15; P = 0.066). No other significant associations were detected with regard to other lipids in models 2–4 (data not shown).
Table 1. Subject characteristics
| ||Duration of MDD symptoms||Healthy controls||Test statistics‡‡||P|
|≥3 years n = 43 (%)||<3 years n = 45 (%)||n = 88 (%)|
|Age (years)||50.56 ± 4.80||49.09 ± 6.63||49.86 ± 8.01||χ2 = 1.15||0.56‡|
|Male||19 (44.2)||20 (44.4)||39 (44.3)||χ2 = 0.001||1.00§|
|Married/living with a partner||27 (62.8)||24 (53.3)||67 (76.1)||χ2 = 7.47||0.02§|
|Daily smoking||17 (39.5)||21 (46.7)||9 (10.2)||χ2 = 24.98||<0.001§|
|Alcohol use ≥2 per week||4 (9.3)||10 (22.2)||14 (15.9)||χ2 = 2.74||0.25§|
|Physical exercise ≥1 per week||25 (59.5)||30 (66.7)||–||χ2 = 0.48||0.49§|
|Lipid-lowering medication||5 (11.6)||7 (15.6)||–||χ2 = 0.29||0.59§|
|Antipsychotic medication||9 (20.9)||12 (26.7)||–||χ2 = 0.40||0.53§|
|Antidepressant medication||41 (95.3)||40 (88.9)||–||–||0.44¶|
|BDI scores||30.12 ± 9.96||25.87 ± 9.79||2.65 ± 3.10||χ2 = 129.27||<0.001‡|
|Overweight (BMI ≥ 25)||34 (79.1)||37 (82.2)||47 (53.4)||χ2 = 14.91||0.001§|
|Total cholesterol (mmol/L)||5.57 ± 1.15||5.48 ± 1.01||5.23 ± 0.98||χ2 = 3.64||0.16‡|
|HDL cholesterol (mmol/L)||1.26 ± 0.36||1.50 ± 0.54||1.52 ± 0.39||F = 8.46||0.004†,††|
|LDL cholesterol (mmol/L)||5.10 ± 1.41||4.67 ± 1.02||3.18 ± 0.89||χ2 = 75.23||<0.001‡|
|Triglycerides (mmol/L)||1.74 ± 1.03||1.55 ± 0.72||1.17 ± 0.54||χ2 = 19.30||<0.001‡|
|TC/HDL-C||4.74 ± 1.55||4.00 ± 1.40||3.64 ± 1.06||χ2 = 17.36||<0.001‡|
|LDL-C/HDL-C||4.45 ± 1.97||3.57 ± 1.70||2.24 ± 0.89||χ2 = 56.87||<0.001‡|
|Non-HDL-C||4.31 ± 1.14||3.97 ± 0.90||3.71 ± 0.94||F = 5.59||0.004††|
Table 2. Likelihood of belonging to the long MDD symptom duration group on multivariate logistic regression
|Married or living with a partner||1.58||0.60–4.12||0.352|
|Symptom severity (high BDI scores)||1.06||1.00–1.11||0.038|
|Overweight (BMI > 25)||0.52||0.16–1.73||0.290|
- Top of page
Patients with a symptom duration ≥3 years had lower HDL-C levels, and higher atherogenic indices, that is, they presented alterations indicating an increased susceptibility to coronary heart disease. Low HDL-C doubled the likelihood for belonging to the long symptom duration group.
Depression-related low-grade tissue inflammation is associated with low HDL-C levels.4 This relationship may be bidirectional. Elevated levels of pro-inflammatory cytokines can lead to lowered HDL-C levels. Interestingly, the subjects in the short symptom duration group had no HDL-C alterations compared with healthy controls. This observation indicates that the systemic anti-inflammatory effects of a high HDL-C level may also act as a buffer against depression-related pronounced low-grade inflammation during the first years of illness.12 The present findings also suggest that this effect may cease during extended periods of depressive symptoms, possibly due to an adverse shift in the balance between anti- and pro-inflammatory cytokines.13
Low HDL-C levels have also been associated with high cortisol levels in depressed, but not in healthy individuals.6 Thus, depression-related hypercortisolemia could partly underlie our observations. Nevertheless, at present there are no data showing whether hypercortisolemia is pronounced in subjects with long-term depression. Further mechanistic and kinetic studies on lipid metabolism are warranted to replicate our preliminary observations and determine the factors underlying the relationship between low HDL-C levels and long symptom duration in MDD.
The data on the MDD and control groups were collected from the same area at the same time, and both groups underwent SCID interviews, providing careful assessment of the study subjects' psychiatric status. Thus, the groups can be considered comparable. Furthermore, the euthymic mental status of the healthy control group had been monitored for the previous 7 years, which is a strength. We were also able to adjust the HDL-C difference between the symptom duration groups for several potential confounders. Thus, it is unlikely that the low levels of HDL-C in the long symptom duration group would only reflect an underlying pattern of adverse lifestyles. The self-reported symptom duration, however, may be susceptible to recall bias. Moreover, the possibility of some of the subjects being on cholesterol-lowering diets may have confounded the observations. Some of the previous data also suggest that depressive subtypes may affect the lipid levels.14 Thus, the lack of adjustments for subtypes may form a potential source of bias.
In conclusion, the present findings suggest that low HDL-C levels are specifically associated with long-term depressive symptomatology. Thus, effective treatment of depression is also warranted in order to prevent adverse depression-related biological alterations. The present findings also suggest a need for further research into the role of HDL-C in the regulation of depression-related inflammatory alterations.