Causal effects of lipid-lowering drug targets on psychiatric disorders: A drug–target Mendelian randomization study

Abstract

Objectives

The pleiotropic effects of lipid-lowering therapies on mental health remain incompletely understood. This study aimed to investigate the causal impact of genetically proxied inhibition of three major lipid-lowering drug targets, 3-hydroxy-3-methylglutaryl-CoA reductase (HMGCR), Niemann-Pick C1-like protein 1 (NPC1L1), and proprotein convertase subtilisin/kexin type 9 (PCSK9), on a spectrum of psychiatric disorders using a drug–target Mendelian randomization approach.

Methods

We used genetic variants located within or near the HMGCR, NPC1L1, and PCSK9 gene regions that are associated with low-density lipoprotein cholesterol levels as proxies for pharmacological inhibition. Summary-level data were obtained from large-scale genome-wide association studies for seven psychiatric outcomes: anorexia nervosa, anxiety, bipolar disorder, major depressive disorder, neuroticism, obsessive compulsive disorder, and schizophrenia. The inverse-variance weighted method was employed as the primary Mendelian randomization approach, supplemented by multiple sensitivity analyses to assess robustness.

Results

Genetically proxied inhibition of HMGCR was associated with an increased risk of major depressive disorder (odds ratio = 1.16; 95% confidence interval: 1.07–1.25; p = 4.5e−04). In contrast, NPC1L1 inhibition was associated with a decreased risk of major depressive disorder (odds ratio = 0.88; 95% confidence interval: 0.84–0.92; p = 8.1e−08). PCSK9 inhibition was significantly associated with an increased risk of major depressive disorder (odds ratio = 1.16; 95% confidence interval: 1.06–1.26; p = 8.2e−04) and bipolar disorder (odds ratio = 1.28; 95% confidence interval: 1.19–1.38; p = 9.4e−12). No significant associations were observed between these targets and the remaining psychiatric outcomes.

Conclusions

This study provides genetic evidence that lipid-lowering drug targets exert distinct effects on psychiatric disorders. These findings highlight the importance of further clinical and mechanistic studies, particularly given the widespread use of lipid-lowering therapies in aging populations who are vulnerable to mental health conditions.

Introduction

Lipid-lowering therapies, particularly statins, ezetimibe, and proprotein convertase subtilisin/kexin type 9 (PCSK9) inhibitors, are widely prescribed to reduce the risk of cardiovascular disease (CVD). ¹ Their potential neuropsychiatric effects have attracted increasing scientific attention. ² Depression is a common psychiatric disorder and the leading contributor to global disability. ²

However, observational studies have reported inconsistent findings regarding the associations between lipid-lowering agents and depression. For example, one meta-analysis suggested that statins reduced the risk of depression by 32%, ³ whereas a large cohort study involving 387,954 individuals reported an increased risk of depression among statin users (hazard ratio (HR) = 1.22; 95% confidence interval (CI): 1.12–1.32). ⁴ These observational findings are subject to limitations, including residual confounding and reverse causality. Moreover, the bidirectional relationship between CVD and depression further complicates interpretation of these associations.^5,6 Statins are among the most commonly prescribed medications in older adults, with nearly one in three adults in the United States using them. ⁷ Given the high prevalence of statin use and psychiatric disorders in aging populations, it is critical to understand the full risk–benefit profile of these agents, particularly in individuals undergoing primary prevention without definite indications for statin initiation.

Mendelian randomization (MR) offers a powerful genetic approach to investigate potential causal relationships between exposures and health outcomes, minimizing confounding and reverse causality.^8,9 Recent MR studies have explored the genetic underpinnings of psychiatric disorders and related traits, providing insights into potential therapeutic targets and biological mechanisms.^10–12 Drug–target MR, a subtype of MR, uses genetic proxies that mimic the pharmacological inhibition of therapeutic targets and has been successfully employed to assess the safety and repurposing potential of medications. In this study, we applied a drug–target MR framework to evaluate whether genetically proxied inhibition of three clinically validated targets, 3-hydroxy-3-methylglutaryl-CoA reductase (HMGCR; the target of statins), Niemann-Pick C1-like protein 1 (NPC1L1; the target of ezetimibe), and PCSK9 (the target of monoclonal antibodies and small interfering ribonucleic acid therapies), is causally associated with the risks of seven major psychiatric disorders: anorexia nervosa, anxiety, bipolar disorder, major depressive disorder (MDD), neuroticism, obsessive compulsive disorder (OCD), and schizophrenia. Understanding these associations is essential for optimizing the use of lipid-lowering agents in populations vulnerable to psychiatric comorbidity.

Methods

Study design and rationale

This study employed a drug–target two-sample MR approach to evaluate the causal effects of lipid-lowering drug targets on psychiatric disorders. Drug–target MR utilizes genetic variants located near or within genes encoding therapeutic targets that are associated with a relevant biomarker, in this case low-density lipoprotein cholesterol (LDL-C), as proxies for pharmacological inhibition. This strategy mimics the biological action of lipid-lowering drugs and provides a natural experiment to infer causality between target modulation and health outcomes, minimizing confounding and reverse causality. Three core assumptions are necessary for MR analysis. 1. The selected genetic variants must be robustly associated with the exposure of interest. 2. The variants should not be associated with confounding variables. 3. The variants should influence the outcome exclusively through the exposure pathway. We focused on three major lipid-lowering targets: HMGCR, NPC1L1, and PCSK9. An overview of the study design is illustrated in Figure 1.

Figure 1.

Overview of the study design. HMGCR: 3-hydroxy-3-methylglutaryl coenzyme A reductase; IVW: inverse-variance weighted; NPC1L1: Niemann-Pick C1-like protein 1; PCSK9: proprotein convertase subtilisin/kexin type 9; SNP: single-nucleotide polymorphism.

Selection of genetic instruments

Genetic instruments were extracted from a large-scale genome-wide association study (GWAS) conducted by the Global Lipids Genetics Consortium, involving over 170,000 individuals of European ancestry. ¹³ Instruments were selected according to the following criteria: (a) single-nucleotide polymorphisms (SNPs) located within ±500 kb of each target gene and significantly associated with LDL-C levels at genome-wide significance (p < 5 × 10⁻⁸); (b) low linkage disequilibrium (r² < 0.30) among selected SNPs to ensure independence; (c) exclusion of palindromic SNPs to avoid strand ambiguity; and (d) definition of the LDL-C–lowering allele as the effect allele, ensuring that all selected SNPs exhibited negative beta coefficients for LDL-C, thereby mimicking pharmacological inhibition. For NPC1L1 with limited number of SNPs, sensitivity analysis was performed by setting p <5 × 10⁻⁶.

Psychiatric outcomes

We investigated the effects of drug–target inhibition on seven psychiatric phenotypes: anorexia nervosa, anxiety, bipolar disorder, MDD, neuroticism, OCD, and schizophrenia. As detailed in Supplementary Table 1, GWAS summary statistics for anorexia nervosa (n = 14,477), bipolar disorder (n = 413,466), MDD (n = 173,005), OCD (n = 33,925), and schizophrenia (n = 320,404) were obtained from the Psychiatric Genomics Consortium (PGC). ¹⁴ Data for neuroticism (n = 160,958) were derived from the Genetics of Personality Consortium (GPC), ¹⁵ and summary statistics for anxiety (n = 484,598) were sourced from the UK biobank. The selected GWAS datasets for the exposure and outcome were procured from independent consortia to minimize potential sample overlap. All datasets included predominantly European participants to minimize population stratification bias.

MR analysis

The primary MR analysis was performed using the inverse-variance weighted (IVW) method under a random-effects model, which provides the most statistically efficient estimates when all instruments are valid. To evaluate the robustness of causal inferences, we conducted complementary sensitivity analyses using the IVW under a fixed-effects model, maximum likelihood, MR-Egger, weighted median, penalized weighted median, and weighted mode methods. Additionally, we conducted multivariable Mendelian randomization (MVMR) analysis to determine whether the observed associations represented direct causal effects. The MVMR approach allows simultaneous adjustment for multiple exposures (body mass index and activity), thereby reducing potential confounding and distinguishing direct versus indirect pathways. Heterogeneity across instruments was assessed using Cochran’s Q statistic, and horizontal pleiotropy was evaluated using the MR-Egger intercept test. ¹⁶ A Bonferroni-corrected significance threshold of p < 2.4e−03 (0.05 divided by 21 comparisons: 3 drug targets × 7 outcomes) was used to determine statistical significance. All analyses were performed using the TwoSampleMR package (version 0.6.17) in R software (version 4.4.1).

Results

Selected SNPs

As shown in Supplementary Table 2, a total of 12 SNPs were selected to proxy HMGCR inhibition, 3 SNPs for NPC1L1, and 14 SNPs for PCSK9. The F-statistic was >10, indicating that all the selected instruments were sufficiently strong and unlikely to introduce substantial bias.

Effect of HMGCR inhibition on psychiatric disorders

Genetically proxied inhibition of HMGCR was significantly associated with an increased risk of MDD (odds ratio (OR) = 1.16; 95% CI: 1.07–1.25; p = 4.5e−04; Figure 2). No significant associations were observed between HMGCR inhibition and the risks of anorexia nervosa, anxiety, bipolar disorder, neuroticism, OCD, or schizophrenia. These null findings were consistent across multiple sensitivity analyses, and no evidence of horizontal pleiotropy was detected (all p > 0.05; Supplementary Table 3). However, moderate heterogeneity was observed for anxiety, bipolar disorder, and schizophrenia.

Figure 2.

Forest plot of the associations between genetically proxied inhibition of HMGCR, NPC1L1, and PCSK9 with psychiatric disorders. CI: confidence interval; HMGCR: 3-hydroxy-3-methylglutaryl coenzyme A reductase; NPC1L1: Niemann-Pick C1-like 1; OR: odds ratio; PCSK9: proprotein convertase subtilisin/kexin type 9; SNP: single-nucleotide polymorphism.

Effect of NPC1L1 inhibition on psychiatric disorders

Genetically proxied inhibition of NPC1L1 was associated with a protective effect on MDD (OR = 0.88; 95% CI: 0.84–0.92; p = 8.1e−08), whereas no significant associations were observed for the remaining six psychiatric outcomes. No evidence of horizontal pleiotropy or heterogeneity was found across all tested outcomes (Supplementary Table 3).

Effect of PCSK9 inhibition on psychiatric disorders

Genetically proxied inhibition of PCSK9 was significantly associated with increased risks of MDD (OR = 1.16; 95% CI: 1.06–1.26; p = 8.2e−04) and bipolar disorder (OR = 1.28; 95% CI: 1.19–1.38; p = 9.4e−12). The positive association between PCSK9 inhibition and MDD was also confirmed in sensitivity analyses, including IVW under a fixed-effects model (OR = 1.16; 95% CI: 1.05–1.27; p = 2.1e−03) and the maximum likelihood method (OR = 1.16; 95% CI: 1.06–1.27; p = 2.0e−03). The positive association between PCSK9 inhibition and bipolar disorder was confirmed using multiple sensitivity analyses, including IVW under a fixed-effects model (OR = 1.28; 95% CI: 1.15–1.43; p = 4.6e−06), maximum likelihood (OR = 1.29; 95% CI: 1.16–1.44; p =3.9e−06), weighted median (OR = 1.26; 95% CI: 1.10–1.46; p = 1.1e−03), penalized weighted median (OR = 1.26; 95% CI: 1.10–1.45; p = 7.1e−04), and MVMR (OR = 1.22; 95% CI: 1.10–1.36; p = 2.0e−05) methods. No significant associations were observed between PCSK9 inhibition and the risks of anorexia nervosa, anxiety, neuroticism, OCD, or schizophrenia. Furthermore, no evidence of pleiotropy or heterogeneity was detected across any of the tested outcomes (Supplementary Table 3). When setting p < 5 × 10⁻⁶ for NPC1L1, similar results were observed (Supplementary Table 4).

A summary of the effects of genetically proxied LDL-C–lowering drug targets on all psychiatric outcomes is presented in Figure 3.

Figure 3.

Heatmap of the associations between genetically proxied inhibition of HMGCR, NPC1L1, and PCSK9 with psychiatric disorders. HMGCR: 3-hydroxy-3-methylglutaryl coenzyme A reductase; NPC1L1: Niemann-Pick C1-like 1; PCSK9: proprotein convertase subtilisin/kexin type 9; SNP: single-nucleotide polymorphism.

Discussion

In this drug–target MR study, we found that genetically proxied inhibition of HMGCR and PCSK9 was associated with an increased risk of MDD, whereas NPC1L1 inhibition demonstrated a protective effect. Additionally, PCSK9 inhibition was associated with a higher risk of bipolar disorder. No significant associations were observed between lipid-lowering drug targets and the other psychiatric outcomes, including anorexia nervosa, anxiety, neuroticism, OCD, or schizophrenia.

Long-term reduction in LDL-C through statin therapy has demonstrated substantial cardiovascular benefits in primary and secondary prevention populations. ¹⁷ Newer lipid-lowering agents, including ezetimibe and PCSK9 inhibitors, have expanded the therapeutic options. Recent studies have suggested that these lipid-lowering agents exert pleiotropic effects beyond cardiovascular risk reduction, including potential impacts on neuropsychiatric health. ¹⁸ Given the bidirectional relationship between CVD and mood disorders, where each can increase the risk of the other,^5,6 it remains unclear whether the observed psychiatric risk is attributable to the underlying CVD or to lipid-lowering therapy itself.

Previous studies have reported inconsistent findings regarding the association between statins and depression. Our MR analysis supports a positive association between HMGCR inhibition and MDD, consistent with a large cohort study involving 387,954 participants that reported an increased risk of depression among statin users. ⁴ Similarly, another MR study found a statistically significant increased risk of depression associated with genetically proxied inhibition of statins (OR = 1.15; 95% CI: 1.04–1.19) and PCSK9 inhibitor (OR = 1.19, 95% CI: 1.1–1.29), after correcting for multiple comparisons. ¹⁹ In contrast, a meta-analysis of 13 observational studies (including 9 cohort, 3 case–control, and 1 cross-sectional study, totaling over 5 million participants across 11 countries) found no significant association between statin use and depression after adjusting for publication bias using trim-and-fill methods (OR = 0.87; 95% CI: 0.74–1.02). ²⁰ However, the included studies exhibited considerable heterogeneity, and the results were inconsistent in subgroups of studies published before 2013. As with all observational research, residual confounding remains a major concern. A meta-analysis conducted by Yatham et al. ²¹ evaluated 10 randomized controlled trials (RCTs) evaluating the effects of statins on depressive symptoms. The overall results suggested a significant reduction in depressive symptoms with statin use (standardized mean difference (SMD) = −0.309; 95% CI: −0.525 to −0.094; p = 0.005). However, subgroup analysis showed that the effect was only significant in individuals with baseline depression (SMD = −0.796; 95% CI: −1.107 to −0.486; p = 0.001), but not in nondepressed individuals (SMD = 0.153; 95% CI: −0.047 to 0.353; p = 0.113). These results suggest that statins improve depressive symptoms in those with preexisting depression but have minimal effect, or even potential risk, in the general population. Notably, most of the included RCTs were short in duration or primarily designed to assess cardiovascular endpoints, possibly underestimating psychiatric side effects. A Swedish national cohort study further suggested heterogeneity among different statins: simvastatin use was associated with a 7% reduction in depression risk, whereas atorvastatin use was linked to an 11% increased risk. ²² These findings imply that only certain statins, particularly lipophilic ones, confer neuroprotective benefits. Additional factors that may influence the neuropsychiatric risk of statins include the intensity of lipid-lowering regimens, baseline patient characteristics (e.g. older age and cognitive impairment), and potential pharmacogenetic interactions. ²³ Notably, RCTs typically evaluate short-term mood changes in patients with hyperlipidemia, whereas MR reflects lifelong HMGCR inhibition in the general population. Differences in study populations, exposure timing, and compensatory biological pathways may explain the observed discrepancies.

Evidence regarding the psychiatric safety of PCSK9 inhibitors remains limited, as these agents represent a relatively new class of lipid-lowering drugs. Most available RCTs are still in early phases and are not adequately powered to detect neuropsychiatric events. Nevertheless, our MR analysis results indicate that genetically proxied PCSK9 inhibition is associated with increased risks of MDD and bipolar disorder, without significant effects on other psychiatric outcomes. A meta-analysis of eight RCTs reported that the use of PCSK9 inhibitors was associated with a more than twofold higher risk of neurocognitive impairment (OR = 2.85; 95% CI: 1.34–6.06). ²⁴ However, the limited number and heterogeneity of existing studies, combined with the difficulty of assessing neuropsychiatric events in traditional trials, highlight the need for more targeted, long-term studies that include mental health outcomes.

The mechanisms underlying the observed associations between HMGCR or PCSK9 inhibition and increased risk of depression remain unclear. One hypothesis is that low circulating cholesterol levels disrupt serotonin neurotransmission, which is implicated in mood regulation. ²⁵ Cholesterol is also a major component of neuronal lipid rafts, which are essential for synaptic vesicle exocytosis. Experimental studies have shown that inhibition of neuronal cholesterol biosynthesis via lovastatin impairs synaptic transmission. ²⁶ However, the protective effect of NPC1L1 inhibition against MDD, in addition to lowering cholesterol, suggests that cholesterol alone does not fully explain the neuropsychiatric effects of these drugs. According to the Genotype-Tissue Expression (GTEx) database, HMGCR and PCSK9 are highly expressed in the brain tissue, whereas NPC1L1 is not highly expressed. ¹⁹ This differential expression pattern may have contributed to the target-specific effects observed in our study. For example, statins have been shown to reduce β-amyloid production, a process implicated in Alzheimer’s disease and depression. ²⁷ Similarly, studies in PCSK9 knockout (PCSK9⁻/⁻) mouse models suggest a critical role for PCSK9 in modulating brain lipid composition and neuroinflammation. ²⁸ In this context, our findings of an association between PCSK9 inhibition and increased bipolar disorder risk is particularly noteworthy and warrants further mechanistic investigation.

Our findings carry important clinical implications. Lipid-lowering therapies are widely prescribed across all age groups and are increasingly initiated for primary prevention in middle-aged and older adults, populations in which psychiatric symptoms commonly emerge or recur. Although statins, ezetimibe, and PCSK9 inhibitors offer substantial cardiovascular protection, clinicians should remain vigilant regarding their potential neuropsychiatric consequences, particularly in individuals with a history of mood disorders. Although our MR analyses provide genetic evidence supporting a potential link between lipid-lowering drug targets and psychiatric traits, these findings should be interpreted with caution. Clinical recommendations, including routine psychiatric screening, require further validation in prospective and interventional studies.

This study has several strengths. First, the use of drug–target MR enables causal inference at the level of therapeutic target inhibition, providing clinically actionable insights. Second, we leveraged large, well-powered GWAS datasets for psychiatric outcomes, which enhanced the reliability and precision of our findings. Third, multiple sensitivity analyses consistently confirmed the robustness of the associations observed. However, several limitations should be acknowledged. First, genetically proxied lifelong target inhibition reflects cumulative exposure throughout the lifespan and may not fully capture the short-term pharmacological effects of drug use initiated in adulthood. Second, psychiatric phenotypes were derived from GWAS conducted primarily among individuals of European ancestry, which may limit generalizability to other populations. Third, our genetic instruments for NPC1L1 were limited in number (n = 3), which may have reduced the statistical power and increased uncertainty in causal estimation. Fourth, due to the use of summary-level data, we were unable to perform stratified analyses by sex, age, baseline lipid levels, or psychiatric history, factors that may modify the associations between lipid-lowering therapies and mental health outcomes. Fifth, although more recent GWAS datasets for certain psychiatric traits have been released, we used publicly available and consortium-validated summary statistics to ensure data reliability, reproducibility, and harmonization with the exposure datasets. Future studies incorporating newer GWAS releases are warranted to confirm the robustness of our findings. Finally, sample size calculation was not performed, and the limited number of samples may have affected the statistical significance of our results. ²⁹

Conclusion

Taken together, genetic evidence suggests that the inhibition of HMGCR and PCSK9 is associated with an increased risk of MDD, whereas inhibition of NPC1L1 exerts a protective effect. Moreover, PCSK9 inhibition was also associated with a higher risk of bipolar disorder in our study. These results provide preliminary insights into the potential neuropsychiatric effects of lipid-lowering therapies and underscore the need to consider cardiovascular and mental health aspects in future clinical research. As this is a hypothesis-generating study, further mechanistic and clinical investigations are warranted to validate these findings.

Supplemental Material

sj-pdf-1-imr-10.1177_03000605261416738 - Supplemental material for Causal effects of lipid-lowering drug targets on psychiatric disorders: A drug–target Mendelian randomization study

Supplemental material, sj-pdf-1-imr-10.1177_03000605261416738 for Causal effects of lipid-lowering drug targets on psychiatric disorders: A drug–target Mendelian randomization study by Yan Wang, Xin Liu and Shuo Huang in Journal of International Medical Research

Supplemental Material

sj-pdf-2-imr-10.1177_03000605261416738 - Supplemental material for Causal effects of lipid-lowering drug targets on psychiatric disorders: A drug–target Mendelian randomization study

Supplemental material, sj-pdf-2-imr-10.1177_03000605261416738 for Causal effects of lipid-lowering drug targets on psychiatric disorders: A drug–target Mendelian randomization study by Yan Wang, Xin Liu and Shuo Huang in Journal of International Medical Research