Abstract
Background Major depressive disorder (MDD) and bipolar disorder (BD) often coexist with metabolic syndrome. Both are linked to increased atherogenicity and a higher risk of cardiovascular diseases.
Objectives This meta-analysis seeks to evaluate the relationship between atherogenic indices and mood disorders (MDD/BD), while identifying the most effective atherogenic biomarkers for mood disorders.
Methods This study adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. We searched electronic databases, including PubMed, Google Scholar, and Web of Science, for articles published up to August 1, 2024.
Results In this meta-analysis, 85 eligible studies (14 on BD and 71 on MDD) were included, covering 70,856 participants: 18,738 patients and 52,118 healthy controls. Patients with mood disorders showed significant increases (p < 0.001) in the Castelli Risk Index 2 (CRI2), Atherogenic Index of Plasma (AIP), and (triglyceride or TG + low-density lipoprotein + very low-density lipoprotein)/(high-density lipoprotein cholesterol or HDL + Apolipoprotein A or ApoA) ratio, but not CRI1 and ApoB/ApoA ratio. Significant lower HDL and lecithin: cholesterol acyltransferase activity, and higher TG levels were observed in mood disordered patients compared with controls. There were no significant differences between MDD and BD patients. Most included studies lacked the most essential information on the inclusion and exclusion of important confounders.
Conclusions AIP is the most effective atherogenicity index for mood disorders. Regular lipid profiling and metabolic syndrome screening are crucial in mood disorders. Early intervention with lipid-lowering therapies is recommended to prevent the worsening of atherogenicity and disease progression.
Introduction
Mood disorders including major depressive disorder (MDD) and bipolar disorder (BD) are acknowledged as significant contributors to global disability rates (Walker, McGee et al. 2015). According to the World Health Organization (WHO), the current global prevalence of MDD is estimated to be 4.4%. Mood disorders are known to exhibit a significant comorbidity with metabolic syndrome (MetS) (de Melo, Nunes et al. 2017). Mood disorders and MetS are both characterized by a variety of shared pathways, such as lipid peroxidation, low-grade inflammation, increased insulin resistance, atherogenicity, and diminished antioxidant levels (de Melo, Nunes et al. 2017). Additionally, there are robust bidirectional relationships between mood disorders, atherosclerosis, and cardio-vascular disorders (CVD) (Maes, Kubera et al. 2011, Maes, Ruckoanich et al. 2011). For instance, mood disorders and MDD have been linked to an increased risk of atherosclerosis, heart infarction, and stroke, while stroke and heart infarction are associated with an increased incidence of MDD (Maes, Ruckoanich et al. 2011). The comorbidity between mood disorders and CVD is once again rooted in a variety of shared pathways, such as inflammatory, oxidative, and nitrosative pathways, as well as increased atherogenicity (Maes, Ruckoanich et al. 2011).
Lowered high-density lipoprotein cholesterol (HDL) levels, which indicate increased atherogenicity, underpin mood disorders, MetS, atherosclerosis, and CVD (Maes, Ruckoanich et al. 2011). Serum HDL-cholesterol levels are lower in MDD than in healthy controls (Maes, Smith et al. 1997). Furthermore, there are reports of abnormalities in the levels of triglycerides (TG), total cholesterol, and low-density lipoprotein (LDL) cholesterol (Maes, Zhou et al. 2024). Similarly, atherogenicity indices may be elevated in individuals with mood disorders, such as the Castelli risk index 1 (i.e. total cholesterol/HDL), Castelli risk index 2 (LDL-cholesterol/HDL cholesterol), and the atherogenicity index of plasma (TG/HDL-cholesterol) (Huang and Chen 2004, Huang 2005, Lehto, Hintikka et al. 2008, Moreira, Jansen et al. 2017, Jirakran, Vasupanrajit et al. 2023, Khalfan, Campisi et al. 2023).
The first paper to demonstrate an elevated Castelli risk index 1 in MDD was published in 1994 (Maes, Delanghe et al. 1994). Vargas-Nunes et al. (2015) and Vargas et al. (2014) reported an elevated Castelli risk index 2 and atherogenic Index of Plasma (AIP) in individuals with mood disorders (Vargas, Nunes et al. 2014, Nunes, Piccoli de Melo et al. 2015). However, a limited number of research papers on mood disorders have actually measured total cholesterol and HDL-cholesterol and reported the Castelli risk indices and the AIP index. Additionally, the Maes’ laboratory has published that precisely establishing differences in atherogenicity profiles between MDD and controls is only possible in MDD patients who do not have MetS (Maes, Zhou et al. 2024). The inclusion of subjects with MetS obscures all associations between atherogenicity profiles and MDD, the severity of depression and suicidal behaviors.
In addition, it is worth considering whether there are other atherogenic indices that could be more targeted towards mood disorders than the Castelli risk index 1 and 2, and the AIP. One potential index is based on a combination of TG, LDL and related lipoproteins (e.g. very low-density lipoprotein, VLDL) / HDL and apolipoprotein (Apo)A (Maes, Zhou et al. 2024). Another potentially relevant index in MDD and mood disorders is the ApoB / ApoA ratio (Maes, Zhou et al. 2024).
It is also important to consider the effects of free cholesterol which is atherogenic and even neurotoxic (Almulla, Thipakorn et al. 2023, Maes, Zhou et al. 2024). HDL effectively eliminates free cholesterol from the body through its activity in the reverse cholesterol transport (RCT) (Glomset, Norum et al. 1975, Maes, Delanghe et al. 1994, Maes, Smith et al. 1997, Jirakran, Vasupanrajit et al. 2023) and, consequently, plays a crucial role in safeguarding tissues from the harmful effects of free cholesterol and oxidative stress. The RCT relies heavily on the activity of lecithin: cholesterol acyltransferase (LCAT), an enzyme that plays a crucial role in converting free cholesterol into cholesterylesters (Almulla, Thipakorn et al. 2023). The latter are encapsulated into the HDL particles and subsequently removed from the circulation by the liver (Almulla, Thipakorn et al. 2023). In addition to LCAT, ApoA is associated with the HDL particle and contributes to its antioxidant and anti-inflammatory functions. Consequently, the heightened atherogenicity and reduced RCT can result in lipid peroxidation, giving rise to neoantigens like oxidized LDL. This, in turn, triggers autoimmune reactions against oxidized LDL and sets off inflammatory processes (Almulla, Thipakorn et al. 2023). These processes are fundamental to the development of both atherosclerosis and mood disorders, including MDD (Almulla, Thipakorn et al. 2023). LCAT activity can be estimated from measurements of total cholesterol and cholesteryl-esters by computing the esterified cholesterol ratio (Maes, Delanghe et al. 1994).
Nevertheless, the available data does not provide any conclusive evidence regarding the atherogenicity biomarker that is most strongly associated with MDD/BD, nor did it establish the clinical value of using one of the lipid ratios to assess heightened atherogenicity in MDD/BD. Therefore, this systematic review and meta-analysis will investigate the potential association between atherogenicity indices and MDD/BD, as well as explore how the presence of MetS and other confounding factors may impact the interpretation of these indices in MDD/BD. Thus, this study aims to investigate the most appropriate atherogenicity marker for psychiatrists to utilize in clinical practice. This marker may help to estimate atherogenicity and the subsequent heightened risk of atherosclerosis, CVD, heart infarction, and stroke.
Materials and methods
To ensure a methodologically robust approach in our research, we adhered to significant frameworks including the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 (Page, McKenzie et al. 2021), and the guidelines for conducting Meta-Analyses of Observational Studies in Epidemiology (Higgins JPT 2019). We analyzed the differences in atherogenicity indices such as the Castelli Risk Index-1 and 2, the AIP, (TG+LDL+VLDL)/(HDL+ApoA), and ApoB/ApoA between mood disorder patients and controls, and between BD and MDD. Additionally, our analysis also included examining the individual’s lipid biomarkers such as total cholesterol, TG, LDL, VLDL, ApoB, HDL, ApoA, and LCAT activity.
Search Strategy
Our methodology for compiling comprehensive data on lipid biomarkers included an extensive search of electronic databases, specifically PubMed, Google Scholar, and Web of Science, from January to August 1, 2024. The search was conducted using predefined keywords and MeSH terms, as listed in Table 1 of the supplementary electronic file (ESF). To ensure robustness of the included data, we also carefully examined the reference lists from selected studies and relevant prior meta-analyses, aiming to include all relevant research in our review.
Eligibility criteria
In this meta-analysis, primary inclusion was reserved for studies published in peer-reviewed journals and authored in English, though we also reviewed grey literature and contributions in Thai, French, Spanish, Turkish, German, Italian, and Arabic. We focused on observational, case-control, and cohort studies with controls and MDD patients diagnosed according to DSM or ICD criteria. Studies eligible for inclusion also included those with baseline biomarker assessments in prospective cohort studies or follow-up analyses. Exclusions applied to animal, genetic, and translational research; studies lacking a control group; those employing non-standard mediums such as saliva, hair, whole blood, and platelet-rich plasma; and articles failing to report mean and standard deviation (SD) or standard error (SE) for biomarkers. Nevertheless, for studies missing mean (SD) or (SE) data, we contacted authors requesting this information. In the absence of a response, we calculated mean (SD) from median values using the approach by Wan et al. (Wan, Wang et al. 2014) and an online Mean Variance Estimation tool and estimated mean (SD) from graphical data using Web Plot Digitizer.
Primary and secondary outcomes
The primary outcome variables of this meta-analysis are the three atherogenicity indices (Castelli risk index 1 and 2, and AIP) in comparison with the (TG+LDL+VLDL)/(HDL+ApoA) ratio and the ApoB/ApoA index in MDD/BD patients versus controls. Additionally, our analysis also included secondary outcomes, namely the individual’s lipid biomarkers such as total cholesterol, TG, LDL, VLDL, ApoB, HDL, ApoA, and LCAT activity.
Screening and data extraction
The initial screening of studies for potential inclusion in our meta-analysis was conducted by KJ and AA, who reviewed the titles and abstracts according to predetermined inclusion criteria, then proceeded to access the full texts of qualifying papers while excluding those not meeting our criteria. Utilizing a specially designed Excel spreadsheet, they catalogued critical information such as author names, study dates, names of lipid biomarkers, their mean values and SD, the number of participants in patient and control groups, and overall sample sizes. This spreadsheet also included details on study design, type of biological sample (e.g., serum, plasma, CSF, brain tissues, blood cells), psychiatric rating scales, disease stage, average ages (mean and SD), sex distribution, and geographical location of each study. In instances of disagreement, the final adjudication was sought from the senior author (MM).
To assess the methodological quality of included studies, we applied the Immunological Confounder Scale (ICS) (Andres-Rodriguez, Borras et al. 2020), which was adapted for atherogenicity research by the senior author (MM). This tool, detailed in Table 2 of the ESF, comprises the Quality Scale and the Redpoints Scale, both of which have been frequently utilized in prior meta-analyses to evaluate studies on immune and tryptophan catabolite data in individuals with affective disorders (Almulla, Thipakorn et al. 2022, Almulla, Thipakorn et al. 2022, Vasupanrajit, Jirakran et al. 2022). The Quality Scale, which assesses factors such as sample size, control of confounding variables, and sampling duration, scores studies from 0 (lower quality) to 10 (higher quality). The Redpoints Scale aims to identify potential biases in atherogenic biomarkers outcomes and study designs by measuring the degree of control over critical confounding factors, with scores ranging from 0 (highest control) to 26 (no control).
Data analysis
In conducting this meta-analysis, we utilized the CMA V4 software and adhered to the PRISMA guidelines, as outlined in Table 3 of the ESF, requiring a minimum of two studies per lipid biomarker. For calculating the Castelli Risk Index 1 and 2, we assumed dependence in comparison between MDD patients and healthy controls and selected the effect sizes positive for total cholesterol and LDL while negative effect size for HDL. Furthermore, the (TG+LDL+VLDL)/(HDL+ApoA) ratio was calculated by selecting positive effect sizes for TG, LDL, and VLDL and negative for HDL and ApoA. Lastly, the directions of the effect sizes were selected positive for ApoB and negative for ApoA. Group analyses were conducted to compare outcomes between MDD and BD, as well as between studies that included participants with MetS and those that excluded them. If no significant differences were found, the groups were combined. We adopted a random-effects model using constrained maximum likelihood for pooling effect sizes, defining statistical significance as a p-value less than 0.05, and expressing effect sizes as standardized mean differences (SMD) with 95% confidence intervals (CIs). SMDs of 0.80, 0.5, and 0.20 were indicative of large, moderate, or small effect sizes, respectively, following Cohen’s classification (Cohen 2013). Heterogeneity was assessed using tau-squared (and secondary by Q, and I2 values), while meta-regression was applied to explore the sources of heterogeneity.
To verify that the effect estimates are robust, leave-one-out sensitivity analyses were performed. The fail-safe N approach, continuity-corrected Kendall tau, and Egger’s regression intercept were used to assess publication bias; the last two methods yielded one-tailed p-values. The trim-and-fill method was used to account for any missing studies and compute modified effect sizes in cases of asymmetry identified by Egger’s test. Additionally, both observed and imputed missing studies were presented using funnel plots, which were used to visualize tiny study effects.
Results
Search results
Using predetermined keywords and MeSH terms, as listed in Table 1 of the ESF, we conducted a systematic search and found 26,593 studies. The PRISMA flow chart presented in Figure 1 gives an overview of the procedure, showing the total number of publications included and excluded. After the first round of screening and eliminating research deemed irrelevant, 25,010 articles were discarded, resulting in a selection of 1,583 studies. 90 research (were judged suitable for inclusion in the meta-analysis after an additional 1,493 publications were eliminated based on our inclusion-exclusion criteria. Furthermore, we omitted 5 studies due to the lack of required data from the analysis. Overall, 85 studies (10 BD, 71 MDD, and 4 including both MDD and BD cohorts) were used to provide data from 70,856 participants, ages ranging from 15 to 78 years, were included in this meta-analysis (Oxenkrug, Branconnier et al. 1983, Bajwa, Asnis et al. 1992, Maes, Delanghe et al. 1994, Olusi and Fido 1996, Maes, Smith et al. 1997, Khalid, Lal et al. 1998, Partonen, Haukka et al. 1999, Sanyal, Chakrabarti et al. 2000, Bilici, Efe et al. 2001, Sevincok, Buyukozturk et al. 2001, Äijänseppä, Kivinen et al. 2002, Ranjekar, Hinge et al. 2003, Ergün, Uguz et al. 2004, Huang and Chen 2004, Karlović, Buljan et al. 2004, Kim and Myint 2004, Nakao and Yano 2004, Huang 2005, Jow, Yang et al. 2006, Sarandol, Sarandol et al. 2006, Onuegbu, Agbedana et al. 2007, Lehto, Hintikka et al. 2008, Cizza, Eskandari et al. 2009, Hamidifard, Fakhari et al. 2009, Kodydková, Vávrová et al. 2009, Sagud, Mihaljevic-Peles et al. 2009, Ahola, Thorn et al. 2010, Das, Malhotra et al. 2010, Heckbert, Rutter et al. 2010, Lehto, Niskanen et al. 2010, Roohafza, Sadeghi et al. 2010, van Reedt Dortland, Giltay et al. 2010, ALİYAZICIOĞLU, DEĞER et al. 2011, Baghai, Varallo-Bedarida et al. 2011, Kotan, Sarandol et al. 2011, Olié, Picot et al. 2011, Ruljancic, Mihanovic et al. 2011, Sadeghi, Roohafza et al. 2011, Sonal Sukreet and Chaturvedi 2011, Hocaoglu, Kural et al. 2012, ERMAN, KARA et al. 2013, Rybka, Kędziora-Kornatowska et al. 2013, Bortolasci, Vargas et al. 2014, Kale, Kale et al. 2014, Palta, Golden et al. 2014, Patra, Khandelwal et al. 2014, Rahiminejad, Moaddab et al. 2014, Scharnholz, Gilles et al. 2014, Vargas, Nunes et al. 2014, Kahl, Schweiger et al. 2015, Liu, Zheng et al. 2015, Nunes, Piccoli de Melo et al. 2015, Ormonde do Carmo, Mendes-Ribeiro et al. 2015, Peng, Xiang et al. 2016, Tunçel Ö, Akbaş et al. 2016, Akgün, Köken et al. 2017, Ekinci and Ekinci 2017, Messaoud, Mensi et al. 2017, Moreira, Jansen et al. 2017, Peng, Zhong et al. 2017, Baghai, Varallo-Bedarida et al. 2018, Enko, Brandmayr et al. 2018, Segoviano-Mendoza, Cárdenas-de la Cruz et al. 2018, Tunçel Ö, Sarısoy et al. 2018, Al-Amarei, Rasheed et al. 2019, Eidan, Al-Harmoosh et al. 2019, Hui, Yin et al. 2019, Péterfalvi, Németh et al. 2019, Su, Li et al. 2019, Wagner, Musenbichler et al. 2019, Karadeniz, Yaman et al. 2020, Zhang, Yang et al. 2020, Guidara, Messedi et al. 2021, Kennedy, Islam et al. 2021, Vaghef-Mehrabani, Izadi et al. 2021, Draghici 2022, Kasak, Ceylan et al. 2022, Shapiro, Kennedy et al. 2022, Wei, Wang et al. 2022, Jirakran, Vasupanrajit et al. 2023, Khalfan, Campisi et al. 2023, Chen, Sun et al. 2024, Maes, Zhou et al. 2024, Qi, Wang et al. 2024).
These participants included 52,118 healthy controls and 18,738 patients (6,622 BD and 12,425 MDD). Turkey, India, and China led the way in terms of the geographical contribution of studies, with 14, 7, and 9 articles, respectively. Additionally, as indicated in ESF, Table 5, the USA, Brazil, Finland, Germany, and Iran each published five research papers, and a wide range of other nations contributed one to four studies each. ESF, Table 5 provides specifics on the quality and redpoint scores determined by the study and expressed as median (min–max).
Primary outcome variables
Atherogenic indices
No significant differences were identified between MDD and BD patients in any of the group analyses; hence, both groups were combined. Table 1 depicts the effect size for the Castelli Risk Index 1 based on 83 studies. Table 2 and ESF, Figure 1 reveal no significant difference in the Castelli Risk Index 1 between MDD/BD patients and healthy controls. Nonetheless, the publication bias analysis (Table 3) uncovered 17 missing studies on the funnel plot’s left side, and incorporating these missing studies shifted the SMD value (-0.088) to a significant decrease.
The effect size for Castelli Risk Index 2 was determined using data from 72 studies. Among these, 9 studies demonstrated CIs entirely below zero, and 18 studies exhibited CIs exclusively above zero. Most of the studies (n=45) had overlapping CIs with varied outcomes; 16 indicated negative SMD values, while 29 documented positive SMD values, as shown in Table 1. Table 2 and Figure 2 indicate that Castelli Risk Index 2 was significantly higher in patients with MDD/BD than healthy controls, albeit the SMD value was modest. There was no publication bias (see Table 3).
We examined data from 76 studies to assess the effect size of the AIP (TG/HDL ratio). This analysis revealed that 6 studies had CIs entirely below zero, while 27 had CIs exclusively above zero. A significant portion of the research, involving 43 studies, showed overlapping CIs with diverse outcomes: 11 yielded negative SMD values, and 32 reported positive SMD values, as elaborated in Table 1. Table 2 and Figure 3 illustrate that the TG/HDL ratio was significantly higher in patients with MDD/BD versus healthy controls. There was no publication bias.
We conducted a meta-analysis of 12 studies to assess the effect size of the ApoB/ApoA ratio. The results showed that two studies had CIs entirely below zero, while four studies had CIs fully above zero. Six studies exhibited overlapping CIs with mixed outcomes: three studies reported negative SMDs, and three reported positive SMDs, as detailed in Table 1. Table 2 and Figure 4 demonstrate that patients with MDD/BD had significantly higher ApoB/ApoA ratios compared to healthy controls. The publication bias analysis in Table 3 identified three missing studies on the right side of the funnel plot. Including these missing studies would increase the effect size to 0.386.
We analyzed data from 78 studies to determine the effect size of the (TG+LDL+VLDL)/(HDL+ApoA) ratio. Our findings show that 5 studies displayed CIs completely below zero, whereas 20 presented CIs above zero. In 53 studies, we found overlapping CIs with varying outcomes; 17 studies documented negative SMD values, and 36 studies reported positive SMD values, as detailed in Table 1. Table 2 and Figure 5 indicated that the (TG+LDL+VLDL)/(HDL+ApoA) ratio was significantly elevated in patients with MDD/BD compared to healthy controls. Publication bias (Table 3) suggested 12 studies missing from the left side of the funnel plot. Upon imputation, these studies result in a reduction of the SMD to non-significant level.
Secondary outcome variables
Total cholesterol
This study analyzed data from 75 studies to assess the effect size of total cholesterol. The analysis revealed that 20 studies had CIs entirely below zero, and 13 had CIs exclusively above zero; 42 studies exhibited overlapping CIs with varied outcomes: 19 yielded negative SMD values, and 23 studies reported positive SMD values (see Table 1). The findings presented in Table 2 and ESF, Figure 2 indicate that total cholesterol levels were not significantly different in patients with MDD/BD compared to healthy controls. Egger’s and Kendall’s tests (Table 3) identified publication bias with 12 missing studies on the left side of the funnel plot; incorporating these studies would decrease the effect size rendering a significant inverse association.
Triglycerides
We evaluated data from 69 studies to determine the effect size of TG as shown in Table 1. This analysis found that 7 studies had CIs entirely below zero, while 25 studies exhibited CIs above zero. Furthermore, 38 studies showed overlapping CIs with diverse outcomes: 9 reported negative SMD values, and 29 documented positive SMD values. Table 2 and ESF, Figure 3 results indicated that the triglyceride levels were significantly elevated in patients with MDD/BD compared to healthy controls.
VLDL
We analyzed data from 9 studies to assess the effect size of the VLDL. As shown in Table 2 and ESF Figure 4, the results indicated no significant difference in VLDL levels between patients with MDD/BD and healthy controls. Additionally, Table 3 did not show evidence of bias in the VLDL findings. The affect size of TG+VLDL was determined by analyzing 69 studies as detailed in Table 1. The results, shown in Table 2 and ESF Figure 5, revealed a significantly elevated TG+VLDL composite score in patients with MDD/BD compared to healthy controls.
LDL
We analyzed data from 59 studies to assess the effect size of LDL levels, as shown in Table 1. The findings, displayed in Table 2 and ESF Figure 6, revealed no significant difference in LDL levels between patients with MDD/BD and healthy controls.
ApoB
In this study, we assessed ApoB levels using data from 12 studies. As presented in Table 2 and ESF Figure 7, the analysis found no significant differences in ApoB levels between patients with MDD and healthy controls. Table 3 highlights one potentially missing study on the left side of the funnel plot. After including this missing study, there was a slight decrease in the SMD, but the result remained non-significant.
HDL
The data from 69 studies was combined to examine the effect size of HDL. Our findings indicated that 25 studies displayed CIs entirely below zero, and 9 showed CIs above zero. A considerable part of the dataset, consisting of 35 studies, presented overlapping CIs with varying results; 23 studies documented negative SMD values, and 12 reported positive SMD values, as detailed in Table 1. HDL levels were significantly lower in MDD/BD patients than in healthy controls, as demonstrated in Table 2 and ESF, Figure 8. Publication bias analysis in Table 3 uncovered 16 missing studies on the left side of the funnel plot. Imputing these studies would further reduce the effect size. Furthermore, group analysis (Table 4) revealed a significant difference (p<0.0001) between studies that included patients with MetS and studies that excluded them.
ApoA
We utilized data from 10 studies to assess the effect size of ApoA. The findings in Table 2 and ESF Figure 9 demonstrated no significant difference in ApoA levels between patients with MDD/BD and healthy controls.
LCAT enzyme activity
This study evaluated data from 3 studies to determine the effect size of LCAT enzyme activity. The analysis showed that two studies had CIs entirely below zero, while one study had overlapping CIs with a negative SMD. As presented in Table 2 and ESF Figure 10, the results indicate a significant reduction in LCAT enzyme activity in patients with MDD/BD compared to healthy controls. Egger’s and Kendall’s tests, as shown in Table 3, revealed no evidence of bias in the findings.
Meta-regression analyses
To identify the specific factors contributing to the observed heterogeneity in studies examining atherogenicity in MDD/BD, a meta-regression analysis was conducted. The findings, detailed in ESF Table 6, indicate that the heterogeneity in atherogenicity indices is significantly influenced by the percentage of Caucasian participants, redpoints scores, MetS, sex, and the status of the index of episode.
Discussion
Atherogenicity indices and mood disorders
The first significant finding of this meta-analysis is that MDD/BD was substantially associated with two frequently employed atherogenicity indices: the AIP and the Castelli risk index 2, whereas there were no significant associations with Castelli risk index 1 and the ApoB/ApoA ratio. In addition, no differences in any of the indices reported here were found between MDD and BD.
The reason for the absence of an association between MDD/BD and the Castelli risk index 1 can be inferred from the analyses of the individual lipids. Specifically, the numerators of the three atherogenic ratios exhibit variations in their associations with MDD/BD, despite the fact that HDL is the denominator. Thus, triglyceride levels (the numerator in the AIP) were substantially and positively associated with MDD/BD, whereas no such changes were found for LDL (numerator in Castelli risk index 2) and total cholesterol (numerator in Castelli risk index 1). In actuality, the AIP index’s superior performance in comparison to the Castelli risk indices 1 and 2 is due to the distinctions in the numerators, as determined by SMD and z values. As a result, AIP may be perceived as the most critical biomarker for MDD/BD among the classical atherogenicity ratios.
Moreover, the heterogeneity assessments indicate that AIP exhibited the lowest heterogeneity (as measured by the τ2 index) in comparison to both Castelli indices and the ApoB/ApoA index. Additionally, clinical investigations that compared the AIP index with the Castelli risk indices concluded that the AIP is a more effective index for MDD/BD. For instance, Nunes et al. (2015) demonstrated that the AIP and Castelli risk index 2 were elevated in MDD/BD (Nunes, Piccoli de Melo et al. 2015). Conversely, individuals with MDD/BD and concomitant tobacco use disorder (TUD) exhibited elevated atherogenicity levels in comparison to the distinct groups of MDD/BD and TUD patients and controls. Additionally, the association between AIP and MDD/BD was more robust than that of the Castelli risk index 2 with MDD/BD (Nunes, Piccoli de Melo et al. 2015). In a separate investigation, the Castelli risk indices were significantly higher in MDD patients than in controls (Vargas, Nunes et al. 2014). However, another study showed that perhaps the ApoB/ApoA ratio was an adequate biomarker for MDD (Maes, Zhou et al. 2024).
Additionally, the present investigation demonstrated that the heterogeneity of both Castelli risk indices was exacerbated by confounding variables, as evidenced by our “red point score.” This suggests that the heterogeneity of these markers is significantly increased when confounders are not adjusted or controlled for. Most importantly, the presence of MetS significantly impacted both Castelli indices. On the other hand, the heterogeneity in the AIP index did not seem to be influenced by red points or the presence of MetS. The latter is important as this systemic review showed that only 7 out of 85 studies excluded subjects with MetS, whilst the remaining studies did not clearly mention exclusion criteria with regards to MetS. Thus, only 7 papers explicitly stated that they excluded individuals with MetS. This is clearly insufficient, although it suggests that even without accounting for MetS and its components, MDD/BD are associated with heightened atherogenicity, as measured with the AIP index.
In fact, a recent study demonstrated that the examination of the associations between MDD and atherogenicity profiles and separate lipids did not yield any significant results in a combined cohort of MDD and MetS patients (Jirakran, Vasupanrajit et al. 2023). Nevertheless, highly significant associations were observed between atherogenicity indices and MDD after individuals with MetS were excluded. Moreover, even indicants of subclinical MetS (when one or more of the latter 5 MetS components are present) may affect lipid levels measured in MDD (Maes, Vasupanrajit et al. 2024).
Alternatively, we anticipated that a composite of atherogenic compounds in the numerator (e.g., TG + LDL + VLDL) and a composite of antiatherogenic compounds in the denominator (HDL + ApoA) may produce superior outcomes. This ratio was highly significantly associated with MDD/BD and showed lower heterogeneity as compared with the other indices.
Nevertheless, it is also important to investigate publication bias in the different atherogenicity markers. Other pro-arguments to use the AIP are that this index did not seem to be influenced by red points, the presence of MetS, or publication bias. Conversely, substantial publication bias was observed in Castelli risk index 1, the ApoB/ApoA ratio, and the TG + LDL + VLDL / HDL + ApoA ratio. Following the imputation of absent values on the left side, a negative correlation was observed between MDD/BD and the Castelli risk index 1. The ratio of TG + LDL + VLDL to HDL + ApoA lost its significance after imputing missing values, but the ApoB/ApoA ratio gained significance.
All in all, the results of these studies indicate that MDD/BD is associated with an increase in atherogenicity, especially when assessed with the AIP index. The absence of adjustment for MetS or its components may obscure the precise associations with the other atherogenicity indices. Consequently, the detection of any associations is contingent upon the number of individuals with MetS who are included in the study. It appears that the most precise method is to exclude subjects with MetS (Maes et al., 2024).
Lipid levels in MDD
Our meta-analysis found that MDD and BD were substantially and negatively associated with HDL cholesterol and LCAT, and positively with TG. Nonetheless, the heterogeneity in those individual markers exceeded that observed with the AIP index. The variability of HDL values was more pronounced than that of the AIP and Castelli indices. The primary obstacle is that HDL is significantly affected by the presence of MetS as detected in our group analysis. As with the AIP index, no significant bias was observed in HDL and triglyceride values. These results show that the AIP is a superior index and that its robustness is contingent upon both its numerator and denominator, which are both linked to MDD/BD.
Our systematic evaluation encompassed a limited number of papers on LCAT activity, which indicated an inverse association between LCAT and MDD. Nevertheless, the heterogeneity is large, although no publication bias was present. Consequently, future research should concentrate on the measurements of free cholesterol, cholesteryl esters, and LCAT activity. Instead of LCAT activity based on the assays of free cholesterol (Maes, Delanghe et al. 1994), it is possible to evaluate LCAT concentrations or use a method to directly measure LCAT activity. Nevertheless, the evaluation of LCAT activity derived from free cholesterol and cholesteryl esters is arguably the most appropriate, as it is based on lipid measurements that have functional properties associated with MDD (Maes, Zhou et al. 2024).
Other confounders
The discovery that there are various MDD/BD subgroups and features that influence atherogenicity is another important confounder that could result in a significant heterogeneity in atherogenicity results. Maes et al. (2024) reported that two subtypes of MDD/BD could be distinguished, namely major dysmood disorder (MDMD) and simple dysmood disorder (SDMD), based on clinical features such as the recurrence of episodes, suicidal behaviors, and the severity of the current episode (Maes, Zhou et al. 2024). The former is the more severe subtype, characterized by a higher recurrence of illness than SDMD, which is the less severe subtype. SDMD is characterized by a low number of episodes or a first episode (Maes, Zhou et al. 2024). The most significant distinctions between the two groups are the immune-inflammatory mediators, T cell activation, growth factors, and atherogenicity profiles (Maes, Rachayon et al. 2022, Maes, Rachayon et al. 2023, Maes, Zhou et al. 2024). In the group of individuals without MetS, it was observed that various atherogenic indices (including ApoB/ApoA) were substantially higher in MDMD than in controls and SDMD (Maes, Zhou et al. 2024). Additionally, the atherogenicity profiles in MDD without MetS were significantly influenced by the recurrence of illness (as measured by the number of episodes and suicidal behaviors) and adverse childhood experiences (Maes, Zhou et al. 2024). The latter study determined that the Research and Diagnostic Algorithmic Rule (RADAR) method, a recently developed diagnostic system, should incorporate ApoB/ApoA and RCT indexes (Maes, Zhou et al. 2024).
It is of the utmost importance to consider and incorporate atherogenicity indices into a diagnostic system for MDD/BD. Firstly, the current body of evidence indicates that there is a strong association between an increased susceptibility to MDD/BD and MetS syndrome or atherogenicity (Nousen, Franco et al. 2013). MetS comorbidity is linked to a more complex affective presentation, a reduced likelihood of recovery, and an increased frequency of episodes and suicide attempts (Fagiolini, Frank et al. 2005, Thomas, Pellicciari et al. 2008, Fries, Pfaffenseller et al. 2012, Grande, Magalhães et al. 2012). Secondly, reduced HDL cholesterol levels in the bloodstream and dysfunctions in the reverse transport of cholesterol can be identified as metabolic risk factors that are shared between mood disorders and CVD, including arteriosclerosis and stroke (Maes, Smith et al. 1997, Maes, Christophe et al. 1999). Third, MDD is characterized by substantial interactions between immune-inflammatory biomarkers and MetS or atherogenicity indices, which result in a more severe phenotype (Maes, Jirakran et al. 2024). In reality, the presence of MetS or atherogenicity in MDD is associated with an aggravation of immune-inflammatory pathways. Additionally, the first episode of SDMD without MetS is distinguished by lowered LCAT activity and increased free cholesterol indicating that this condition is characterized by a pre-atherogenic state even in the absence of MetS (Maes, Vasupanrajit et al. 2024).
Limitations
The correct evaluation of atherogenicity biomarkers in mood disorders is fraught with numerous sources of heterogeneity, including MetS, race, and the phase of the index episode. Most studies failed to account for MetS or its components. The phenotypic features of mood disorders, such as recurrence of illness and distinct subtypes, may be the most significant confounders, in addition to MetS.
Conclusions
The findings of this investigation suggest that mood disorders are linked to elevated levels of atherogenicity. The heterogeneity of the data and the presence of bias somewhat impede the selection of the most appropriate atherogenicity biomarker. However, it may seem that the AIP index may be the most effective atherogenic biomarker for MDD/BD. Future research should investigate whether the application of LCAT, free cholesterol, ApoA, or ApoB, measurements can enhance the relevance of atherogenicity biomarkers in the context of MDD.
In any event, the findings of this paper indicate that it is imperative to conduct routine MetS screenings for mood disorder patients. Psychiatrists must consistently assess atherogenicity (by assessing total cholesterol, LDL, HDL, TG, free cholesterol, LCAT activity) and determine the presence of MetS utilizing the American Heart Association criteria (American Heart Association 2023). Patients with MetS or hyperlipidemia should receive treatment for these disorders, since they may exacerbate the phenotype and stage of mood disorders. In the absence of MetS, psychiatrists should assess the AIP as TG/HDL cholesterol. Lipid-lowering therapies (e.g. statins, diets, omega-3 supplements) should be commenced to avert the exacerbation of atherogenicity when the AIP index is increased. It appears that anti-atherogenic therapies are essential to avert recurrence of mood disorders, the exacerbation of the phenome of depression, and suicidal tendencies, all of which are associated with heightened atherogenicity (Maes, Zhou et al. 2024). Similarly, individuals experiencing their first episode of MDD should be assessed for preclinical MetS through the evaluation of LCAT, free cholesterol, and ApoE (Maes, Zhou et al. 2024).
Author Declarations
Ethical approval and consent to participate
Not applicable.
Consent for publication
Not applicable.
Availability of data and materials
The dataset generated during and/or analyzed during the current study will be available from MM upon reasonable request and once the dataset has been fully exploited by the authors.
Competing interests
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Funding
This work was supported by the Ratchadapiseksompotch Fund, Graduate Affairs, Faculty of Medicine, Chulalongkorn University (Grant number GA64/21), a grant from CU Graduate School Thesis Grant, and Chulalongkorn University Graduate Scholarship to Commemorate the 72nd Anniversary of His Majesty King Bhumibol Adulyadej to KJ, For the Thailand Science research and Innovation fund Chulalongkorn University, MDCU (RA66/016) to MM
Author’s contributions
The work was designed by MM and KJ. Data were collected by KJ, TJ,AV and PJ. Statistical analyses were performed by MM and AA. All authors revised and approved the final draft.
Data Availability
The work was designed by MM and KJ. Data were collected by KJ, TJ,AV and PJ. Statistical analyses were performed by MM and AA. All authors revised and approved the final draft.
Acknowledgements
We are grateful to all participants who participated in this study, nurses, and staff at department of Psychiatry King Chulalongkorn Memorial Hospital, the Thai Red Cross Society, who helped recruit our participants, and Center for Medical Diagnostic Laboratories Faculty of Medicine Chulalongkorn University/King Chulalongkorn Memorial Hospital. This work was supported by the grant from Ratchadapiseksompotch Fund, Graduate Affairs, Faculty of Medicine, Chulalongkorn University (grant numbers RA64/021), Bangkok, Thailand. KJ scholarship from CU Graduate School Thesis Grant and the scholarship from Graduate School, Chulalongkorn University to Commemorate the 72nd anniversary of his Majesty King Bhumibol Adulyadej is gratefully acknowledged.
Footnotes
Emails: ket.kett{at}hotmail.com; Abbass.chem.almulla1991{at}gmail.com, thapanee.ja{at}up.ac.th, asara.vasu{at}gmail.com, Chavit.tun{at}gmail.com, psynjansem{at}docchula.com, stojanovpisevski{at}gmail.com, elizabet.dzhambazova{at}mu-plovdiv.bg
↵# Joined first author
References
