Association between BMI and dyslipidemia in patients with severe mental disorders: An electronic health record study

Abstract

Background

It's well known that obesity is a significant risk factor for dyslipidemia in the general population, but research related to individuals with mental disorders remains scarce. This study aimed to investigate the association between body mass index (BMI) and dyslipidemia among adult patients with severe mental disorders (SMD).

Methods

We implemented a cross-sectional study based on electronic health records. A total of 1145 adult SMD patients who underwent health examinations in 2022 were selected as study subjects. The multivariate logistic regression was used to explore association between BMI and dyslipidemia (including specific lipid abnormalities), and the exposure-response curves were plotted for whole population and subgroups.

Results

Among the study subjects, 574 SMD patients had dyslipidemia. Compared with patients of normal weight, underweight was negatively associated with dyslipidemia (OR = 0.47, 95% CI: 0.24–0.94), while overweight (OR = 1.75, 95% CI: 1.33–2.30) and obesity (OR = 2.26, 95% CI: 1.59–3.21) were positively associated with dyslipidemia. The OR per 1.0 kg/m² increases in BMI was 1.11 (95% CI: 1.08–1.15). The exposure-response curves indicated that associations between BMI and dyslipidemia generally followed an S-shaped curves. Across different subgroups, the association exhibited variations, with greater strength observed among females, individuals aged ≥60 years, and those with hypertension, diabetes, and mental retardation accompanied by mental disorders. Associations between BMI and specific blood lipid abnormalities varied across populations. However, there were no significant interactions between all subgroup factors and BMI.

Conclusions

In the management of SMD patients, it is crucial to strengthen weight control and implement targeted interventions based on individual characteristics and specific types of dyslipidemia.

Introduction

Dyslipidemia is the one of most essential factors leading to atherosclerotic cardiovascular disease.^1,2 Patients with severe mental disorders (SMD) also remained a huge disease burden attributed to dyslipidemia. In China, the prevalence of dyslipidemia among inpatients with mental disorders had reached 14.19% with an upward trend year by year. ³ Among them, the prevalence among patients with schizophrenia was 18.19%, while in some regions, the prevalence among patients with major depressive disorder could even exceed 70%.^3,4 Meanwhile, the prevalence of dyslipidemia varied from populations with individual characteristics, like sex, age, education levels, marital status, the presence of diabetes and hypertension, and types of mental disorders. ³ However, no consistent results were found in different studies.^3–5

In addition, obesity was confirmed as an independent risk factor for cardiovascular disease.^6,7 It was well-known that body mass index (BMI) was the primary indicator for assessing an individual's degree of obesity. ⁸ Meanwhile, previous studies demonstrated that individuals with a higher BMI in the general population had a higher risk for dyslipidemia,^9,10 but there was limited evidence that the similar association in the population with SMD. Although both studies focused on patients with major depressive disorder, the association between BMI and dyslipidemia was entirely opposite.^4,5

Given the considerable inconsistency observed in previous studies, we aimed to explore the association between BMI and dyslipidemia in our local population with SMD and to examine whether this association varies across subgroups factors such as age, sex, the presence of diabetes and hypertension, and SMD types. In this study, we conducted a cross-sectional research design, matching the health examination records of 2022 for all patients with SMD under management. The main outcome variables include dyslipidemia and specific lipid abnormalities. BMI was treated as both categorical and continuous variables to explore its association with dyslipidemia, and we also investigated the exposure-response associations of both in different subgroups.

Methods

Participants

We enrolled patients with SMD who were managed and treated within the administrative region. All data with patients were originated from the Regional Mental Health Information Management System and the Universal Health Information Platform. Inclusion criteria were as follows: (1) Disease diagnoses included six categories: schizophrenia, schizoaffective disorder, paranoid psychosis, bipolar (affective) disorder, psychiatric disorders associated with epilepsy, and mental retardation accompanied by mental disorders; (2) The initial onset of the disease was on or before 31 December 2022; (3) At least one health examination was conducted between 1 January 2022, and 31 December 2022; (4) The age at the time of the last health examination in 2022 was ≥18 years. Exclusion criteria included: (1) Disease diagnoses were solely other types of mental disorders (such as vascular dementia, senile dementia, etc.) or the disease diagnosis was missing; (2) Critical information such as height, weight, blood glucose or diabetes diagnosis, blood pressure or hypertension diagnosis, blood lipids or dyslipidemia diagnosis were missing in the health examination records; (3) Duplicate records. Ultimately, 1145 individuals were included in this study (Figure 1). The management of patients with SMD and their health examinations followed the principle of informed consent.

Figure 1.

Flow diagram for study population selection.

EHR: electronic health record.

Original sources of electronic health record

According to national SMD-related management and treatment requirements, primary healthcare facilities established resident electronic health records and conduct follow-up management for SMD patients diagnosed by professional mental health institutions within their jurisdictions.^11,12 The electronic health record mainly included general demographic information, informed consent for management services, the initial onset time of the illness, diagnosis, and treatment status. With the consent of the guardian and/or the patient themselves, primary healthcare facilities conducted health examinations for the patients, which include physical examinations (such as height, weight, and blood pressure) and laboratory tests (such as blood routine tests, blood lipids, and blood glucose levels).

Definitions of some diseases and abnormal conditions

In this study, dyslipidemia was defined as having a total cholesterol (TC) level ≥6.2 mmol/L and/or a low-density lipoprotein cholesterol (LDL-C) level ≥4.1 mmol/L and/or a high-density lipoprotein cholesterol (HDL-C) level <1.0 mmol/L and/or a triglyceride (TG) level ≥2.3 mmol/L and/or a previous diagnosis of dyslipidemia. ¹ The values of BMI (kg/m²) were classified into four categories: underweight (BMI < 18.5), normal weight (18.5 ≤ BMI < 24.0), overweight (24.0 ≤ BMI < 28.0), and obesity (BMI > 28.0). ⁸ Hypertension was defined as having a systolic blood pressure ≥140 mmHg and/or a diastolic blood pressure ≥90 mmHg and/or a previous diagnosis of hypertension. ¹³ Diabetes mellitus was defined as having a fasting plasma glucose level ≥7.0 mmol/L and/or a previous diagnosis of diabetes. ¹⁴

Statistical analysis

Based on the dyslipidemia definition, we divided the whole population into two groups: dyslipidemia and non-dyslipidemia. Categorical data and quantitative data were statistically described using counts (proportion) and medians (with interquartile ranges), respectively. Group differences between both types of data were conducted with Fisher's exact test and the Wilcoxon's rank-sum test, respectively. Generalized additive models combined with the unconditional logistic regression was used to explore the association between BMI and dyslipidemia, with categorical BMI and continuous BMI values serving as independent variables. Odds ratio (OR) and 95% confidence interval (95% CIs) were assessed for the association strength. For the both forms of BMI, ORs were obtained for different BMI categories with normal weight as the reference group (for categorical BMI) and for per 1.0 kg/m² increases in BMI (for continuous BMI). Natural cubic spline functions were also used to fit exposure-response association between BMI and dyslipidemia through the R package - dlnm. Subgroup analyses were conducted by sex, age, diseases status, SMD types, and specific blood lipid abnormalities (elevated TC, TG, and LDL-C levels, and decreased HDL-C levels). Due to the small number of participants with certain types of mental disorders, this study conducted subgroup analyses only for schizophrenia, mental retardation accompanied by mental disorders, and bipolar (affective) disorder. The interaction between subgroup factor and BMI was determined by using Wald's test. For sensitivity analysis, some covariates were adjusted, such as age, sex, education level, marital status, economic status, and diseases status. All statistical analyses were conducted using Excel and R version 4.4.1, with a significance level of 0.05 (two-tailed).

Results

Overview of participants

Totally, there were 1145 SMD patients enrolled in this study, including 516 males and 629 females, with the overall median ages of 61.94 years. Among the study participants, the three most prevalent types of SMD were schizophrenia, mental retardation accompanied by mental disorders, and bipolar (affective) disorder, accounting for 47.07%, 24.02%, and 17.73% of cases, respectively. All other types each comprised fewer than 100 cases (<10%). The education levels were mainly at the primary school and below, and the proportion decreased with higher education levels. Most of individuals were in marriage (57.73%) and came from non-poor families (80.35%). The median BMI value was 24.03 kg/m² in the whole population. The proportions of four BMI categories were 45.50% (normal-weight), 33.36% (overweight), 17.03% (obese), and 4.10% (underweight). There were 48.47% and 19.91% of participants suffering from hypertension and diabetes mellitus, respectively. Among the study participants, 50.13% were found to have dyslipidemia. Regarding specific abnormal components, 11.53% had elevated TC, 18.78% had elevated TG, 10.31% had elevated LDL-C, and 15.28% had decreased HDL-C. Individuals with varying body weights demonstrated significantly different dyslipidemia prevalence rates, while socio-demographic differences were observed in LDL-C and HDL-C abnormality rates ( Table 1 ).

Table 1.

Characteristics of the study population.

Characteristics	Overall [n(proportion%)]	Dyslipidemia		Elevated TC			Elevated TG			Elevated LDL-C		Decreased HDL-C
		Prevalence (%)	P	Prevalence (%)	P		Prevalence (%)	P		Prevalence (%)	P	Prevalence (%)	P
Whole population	1145 (100)	574 (50.13)		132 (11.53)			215 (18.78)			118 (10.31)		175 (15.28)
Sex			0.096		0.007			0.323			0.019		<0.001
Male	516 (45.07)	273 (52.91)		45 (8.72)			90 (17.44)			41 (7.95)		112 (21.71)
Female	629 (54.93)	301 (47.85)		87 (13.83)			125 (19.87)			77 (12.24)		63 (10.02)
Age (years old) a	61.94 (52.15, 70.40)	60.31 (52.57, 70.05)	0.530	59.44 (53.66, 67.81)	0.386		58.47 (50.79, 68.86)	0.035		57.99 (52.77, 66.82)	0.081	60.99 (53.46, 70.85)	0.651
Age group			0.237		0.355			0.019			0.041		0.681
18–59 years old	544 (47.51)	283 (52.02)		68 (12.50)			118 (21.69)			67 (12.32)		86 (15.81)
≥60 years old	601 (52.49)	291 (48.42)		64 (10.64)			97 (16.14)			51 (8.49)		89 (14.81)
Education levels			0.491		0.084			0.188			0.02		0.94
Primary school and below	695 (60.70)	350 (50.36)		91 (13.09)			118 (16.98)			80 (11.51)		104 (14.96)
Junior middle school	325 (28.38)	156 (48.00)		26 (8.00)			71 (21.85)			22 (6.77)		51 (15.69)
High middle school	97 (8.47)	55 (56.70)		13 (13.40)			22 (22.68)			15 (15.46)		15 (15.46)
College and above	28 (2.45)	13 (46.43)		2 (7.14)			4 (14.29)		1 (3.57)			5 (17.86)
Marital			0.054			0.279		0.499			0.041		0.015
Unmarried	308 (26.90)	172 (55.84)		43 (13.96)			64 (20.78)		43 (13.96)			63 (20.45)
Married	661 (57.73)	314 (47.50)		69 (10.44)			117 (17.70)		62 (9.38)			90 (13.62)
Divorced/Widowed/Other	176 (15.37)	88 (50.00)		20 (11.36)			34 (19.32)		13 (7.39)			22 (12.50)
Economic status			0.458			0.063		0.924			0.020		0.050
Non-poor	920 (80.35)	456 (49.57)		98 (10.65)			172 (18.70)		85 (9.24)			131 (14.24)
Poor	225 (19.65)	118 (52.44)		34 (15.11)			43 (19.11)		33 (14.67)			44 (19.56)
BMI (kg/m2)a	24.03 (21.65, 26.72)	24.94 (22.41, 27.42)	<0.001	25.55 (22.61, 28.08)		0.002	25.72 (23.39, 28.34)	<0.001	25.47 (22.66, 27.47)		0.002	24.22 (22.36, 27.04)	0.071
BMI classification			<0.001			0.007		<0.001			0.024		0.033
Underweight	47 (4.10)	12 (25.53)		6 (12.77)			2 (4.26)		4 (8.51)			2 (4.26)
Normal weight	521 (45.50)	223 (42.80)		44 (8.45)			63 (12.09)		39 (7.49)			76 (14.59)
Overweight	382 (33.36)	217 (56.81)		48 (12.57)			92 (24.08)		49 (12.83)			71 (18.59)
Obesity	195 (17.03)	122 (62.56)		34 (17.44)			58 (29.74)		26 (13.33)			26 (13.33)
Hypertension			0.002			0.268		0.058			0.145		0.870
Without	590 (51.53)	269 (45.59)		62 (10.51)			98 (16.61)		53 (8.98)			89 (15.08)
With	555 (48.47)	305 (54.95)		70 (12.61)			117 (21.08)		65 (11.71)			86 (15.50)
Diabetes mellitus			<0.001			0.202		<0.001			0.051		0.150
Without	917 (80.09)	436 (47.55)		100 (10.91)			152 (16.58)		86 (9.38)			133 (14.50)
With	228 (19.91)	138 (60.53)		32 (14.04)			63 (27.63)		32 (14.04)			42 (18.42)
SMD types			0.004			0.915		0.002			0.494		0.596
Schizophrenia	539 (47.07)	304 (56.40)		62 (11.50)			131 (24.30)		5 (0.93)			90 (16.70)
Mental retardation accompanied by mental disorders	275 (24.02)	122 (44.36)		36 (13.09)			36 (13.09)		32 (11.64)			38 (13.82)
Bipolar (affective) disorder	203 (17.73)	94 (46.31)		19 (9.36)			27 (13.30)		15 (7.39)			31 (15.27)
Psychiatric disorders associated with epilepsy	73 (6.38)	30 (41.10)		9 (12.33)			14 (19.18)		6 (8.22)			7 (9.59)
Schizoaffective disorder	17 (1.48)	8 (47.06)		2 (11.76)			2 (11.76)		8 (47.06)			4 (2.29)
Paranoid psychosis	5 (0.44)	1 (20.00)		0 (0)			0 (0)		11 (220.00)			0 (0)
Multi-SMDs	33 (2.88)	15 (45.45)		4 (12.12)			5 (15.15)		5 (15.15)			5 (15.15)

TC: total cholesterol; TG: triglyceride; LDL-C: low-density lipoprotein cholesterol; HDL-C: high-density lipoprotein cholesterol; SMD: severe mental disorder.

^aAge and BMI are characterized with median and quantiles.

^aBold values indicated significant differences among groups.

Association between BMI and dyslipidemia

After adjusting for age, sex, educational level, marital status, economic status, and the presence of diabetes and hypertension (Model 5), compared to the normal weight group, underweight was negatively associated with dyslipidemia (OR = 0.47, 95% CI: 0.24–0.94), while overweight (OR = 1.75, 95% CI: 1.33–2.30) and obesity (OR = 2.26, 95% CI: 1.59–3.21) were positively associated with dyslipidemia. Similarly, BMI per 1.0 kg/m² increases was positively associated with dyslipidemia (OR = 1.11, 95% CI: 1.08–1.15). These associations remained relatively stable after adjusting for different combinations of factors (Models 1–5) and were all statistically significant (P < 0.05) ( Table 2 ).

Table 2.

Association between BMI and dyslipidemia from logistic regression results [OR (95% CI)].

Variable	Model 1	Model 2	Model 3	Model 4	Model 5
BMI classification
Underweight	0.46(0.23–0.90)	0.48(0.24–0.95)	0.46(0.23–0.91)	0.49(0.25–0.98)	0.47(0.24–0.94)
Normal weight	Reference	Reference	Reference	Reference	Reference
Overweight	1.76(1.35–2.29)	1.77(1.35–2.31)	1.83(1.39–2.40)	1.69(1.29–2.22)	1.75(1.33–2.30)
Obesity	2.23(1.59–3.13)	2.34(1.66–3.30)	2.42(1.71–3.42)	2.19(1.55–3.10)	2.26(1.59–3.21)
BMI per 1.0 kg/m2 increases	1.11(1.08–1.15)	1.12(1.08–1.15)	1.12(1.08–1.16)	1.11(1.07–1.14)	1.11(1.08–1.15)

Note: Model 1 represented a univariate regression model. Model 2 adjusted for age and sex, with age adjusted using a natural cubic spline function (with 3 degrees of freedom). Model 3 further adjusted for education level, marital status, and economic status based on Model 2. Model 4 adjusted for diabetes mellitus and hypertension status based on Model 2. Model 5 adjusted for all other factors in both Model 3 and Model 4. All associations presented in the table were statistically significant (all P-values < 0.05).

Exposure-response association

The exposure-response curves for BMI and dyslipidemia in SMD patients and different subgroups, with BMI at 24 kg/m² serving as the reference level, are shown in Figure 2. Among the whole population, the association exhibited an S-shaped curve, with the OR value gradually increasing as BMI rose and stabilizing from 32.5 kg/m² (OR = 1.80, 95% CI: 1.20–2.70) onwards. The exposure-response curves among different sex subgroups were similar to the overall trend, but the association strength between higher BMI (>24 kg/m²) and dyslipidemia was greater in female populations compared to the male, with a peak at 34.5 kg/m² (OR = 2.23, 95% CI: 1.05–4.72). In individuals aged 60 and above, the association strength at lower BMI was slightly lower than that in the group aged 18–59 years, while the association strength at higher BMI was similar to or greater than that in the latter. Compared to individuals without diabetes, the association between BMI and dyslipidemia in diabetic populations showed an upward trend as BMI increased. Similar results were observed in hypertensive populations compared to those without hypertension, but with a slightly smaller magnitude of change. Among different types of mental disorders, a marked upward trend in the association strength at higher BMI levels was found in populations with mental retardation accompanied by mental disorders, while no such trend did appear in populations with schizophrenia or bipolar disorder. Moreover, the associations in the latter two populations were not significant at higher BMI levels. In addition, there were no significant interaction between all subgroup factor and BMI (P-value > 0.05).

Figure 2.

Exposure-response curves for BMI and dyslipidemia in the overall population and different subgroups of patients with severe mental disorders.

Note: Models were adjusted for age, sex, education level, marital status, economic status, and the presence of diabetes and hypertension (excluding the subgroup-specific factors). The exposure-response curves for BMI and dyslipidemia were fitted using a natural cubic spline function with 3 degrees of freedom.

Subgroups analysis by lipid components

Compared to normal weight group, abnormal lipid levels were not less or more prevalent in underweight individuals. In addition, overweight and obesity were positively associated with elevated TC, TG, and LDL-C levels. Associations between overweight, obesity, and elevated LDL-C level had similar strength, but there was incremental strength from overweight to obese individuals on elevated TC and TG levels. Besides, there were no significant associations between BMI classifications and decreased HLD-C levels (Figure 3).

Figure 3.

Association between BMI classification and abnormal lipid levels.

Note: Models were adjusted for age, sex, education level, marital status, economic status, and the presence of diabetes and hypertension. TC: total cholesterol; TG: triglyceride; LDL-C: low-density lipoprotein cholesterol; HDL-C: high-density lipoprotein cholesterol.

Consistent with BMI-dyslipidemia association, BMI per 1.0 kg/m² increases were positively associated with abnormal lipid levels for all lipid components. The strongest association was observed in BMI-elevated TG levels (OR = 1.13, 95% CI: 1.09–1.18) (Figure 4).

Figure 4.

Association between BMI per 1.0 kg/m² increases and abnormal lipid levels.

Note: Models were adjusted for age, sex, education level, marital status, economic status, and the presence of diabetes and hypertension. TC: total cholesterol; TG: triglyceride; LDL-C: low-density lipoprotein cholesterol; HDL-C: high-density lipoprotein cholesterol.

The J- or U-shaped curves appeared among exposure-response associations between BMI and all types of abnormal lipid levels. In marked contrast to its association with dyslipidemia, the association strength between higher BMI and elevated TC levels was greater in the group aged 18–59 years and in males (Figure S1). In addition, the association between higher BMI and elevated TG levels was stronger among non-hypertension group and individuals with bipolar disorder (Figure S2). Furthermore, the association strength between higher BMI and elevated LDL-C levels was stronger in the group aged 18–59 years and among individuals without hypertension, although weaker than the association between BMI and elevated TG levels (Figure S3). Specially, the association strength between BMI and decreased HDL-C levels intensified with increasing BMI; however, the magnitude of change was minimal starting from a BMI of 24 kg/m². Significant increases were only apparent in non-hypertension individuals and those aged ≥60 years (Figure S4). Similar to associations between BMI and dyslipidemia, there were no significant interaction among that between BMI and specific lipid abnormalities.

Discussions

In this study, we analyzed the association between BMI and dyslipidemia among adult patients with SMD based on electronic health records. The results showed that compared with individuals of normal weight, underweight and overweight/obesity were respectively associated with a negative and positive correlation with dyslipidemia. Similar associations had been performed in previous studies of populations including those with schizophrenia, hypertension, diabetes mellitus, and lactating women,^15–18 but there were various strengths of association, which might be closely related to differences in individual levels of physical activity, dietary behaviors, smoking habits, and external interventions among different populations. ¹⁹ The abnormal lipid metabolism associated with obesity was often caused by insulin resistance, involving complex changes including inflammatory responses, neural conduction, and cellular homeostasis. ²⁰ On the one hand, insulin resistance enhanced lipid breakdown in adipose tissue, and the large amounts of free fatty acids entering the circulatory system were synthesized into new lipids and very low-density lipoprotein in the liver, ultimately leading to hyperlipidemia. ²¹ On the other hand, weakened clearance of postprandial chylomicrons and their remnants favored the enrichment of triglycerides in certain types of lipoprotein particles, subsequently reducing the content of HDL-C in the plasma. ¹⁹

The overall exposure-response association between BMI and dyslipidemia exhibited an S-shaped curve, with the strength of association stabilizing after BMI reached a certain level, suggesting potential underlying adaptive mechanisms. Although some studies had shown that there was a nonlinear or linear exposure-response association between BMI and dyslipidemia, these were mostly limited to non-specific populations.9,10 Further subgroup analyses revealed stronger associations in corresponding subgroups among women, individuals aged 60 and above, those with hypertension, diabetes mellitus, and mental retardation accompanied by mental disorders. The differences observed in sex and age subgroups were inconsistent with the results from general populations in previous studies.^9,10 In addition, a network meta-analysis suggested that various anti-psychotics could cause different degrees of increases in weight, BMI, some blood lipid, and blood glucose indicators. ²² Moreover, there were differences in the absorption, distribution, metabolism, and excretion of anti-psychotics among different ages and sex,^23,24 which were contributed to the variations in the strengths of association between BMI and dyslipidemia observed in different studies. Besides, some studies had shown associations between hypertension and diabetes mellitus with dyslipidemia.^25,26 A study from the China Cardiometabolic Disease and Cancer Cohort indicated that 24.4% of the risk of diabetes in the general Chinese population could be attributed to insulin resistance, and 12.4% was attributed to pancreatic beta-cell dysfunction, ²⁷ with the former being one of the main causes of dyslipidemia. There was an interaction between hypertension and dyslipidemia. Hypertension could be caused by endothelial dysfunction resulting from dyslipidemia, while dyslipidemia was also exacerbated by multi-organ functional changes caused by hypertension. In terms of differences in the strength of association among different types of mental disorders, it may be related to anti-psychotic medication use and adherence, which required deep consideration in future research. Also, there were different links between BMI and specific lipid abnormalities among various populations. Difference of response to lipid metabolism and dietary habits may result in these discrepancies. Meanwhile, there were no significant difference by subgroups. This may suggest that these factors do not modify the overall association between BMI and dyslipidemia. Alternatively, it could be an artifact caused by insufficient sample sizes at various BMI levels, leading to overly wide CIs.

There were several limitations in this study. Firstly, the participants were limited to SMD patients who underwent health examinations within a specific year, and the final sample included in the study may not be entirely consistent with the overall population. Secondly, the laboratory test results for blood lipid indicators among the study subjects could be influenced by the timing of the health examination, potentially leading to misclassification of blood lipid abnormalities. Thirdly, this study did not control for the patients’ medication use due to missing data. Additionally, the inherent limitation of cross-sectional study design, its inability to establish temporal sequence, prevents confirmation of a causal relationship between BMI and dyslipidemia, so large-scale, high-quality longitudinal studies are still needed to substantiate this association.

Conclusion

Overall, among adult patients with SMD across different groups, there were positive associations between increased BMI and dyslipidemia. This suggested that during the follow-up management of SMD patients, weight management should be specifically incorporated based on individual patient characteristics, in order to reduce the disease burden potentially caused by dyslipidemia.