DOI: https://doi.org/10.21203/rs.3.rs-2822676/v1
The objective of this study was to investigate the causal relationship between cholesterol-lowering therapy and Alzheimer's disease (AD) using Mendelian Randomization (MR) with two sets of genetic instruments derived from UK Biobank, GLGC, and GWAS ATLAS. Instrumental variables were selected based on SNPs that were significantly associated with lipid-lowering drugs or targets, but not with outcome or confounding factors. The primary analysis was conducted using inverse variance weighted (IVW), MR-PRESSO, WM. Cochran Q, and MR pleiotropy tests to assess heterogeneity or pleiotropy. The results revealed that cholesterol-lowering drugs did not show a significant effect on AD risk with IVW (Atorvastatin: OR = 0.943, 95% CI = 0.612–1.453, p = 0.789; Pravastatin: OR = 6.857, 95% CI = 0.514–90.864, p = 0.144; Rosuvastatin: OR = 2.466, 95% CI = 0.333–18.278, p = 0.377; Simvastatin: OR = 1.138, 95% CI = 0.976–1.328, p = 0.098; Ezetimibe: OR = 1.292, 95% CI = 0.239-6,969, p = 0.766). Further multivariable and target MR analyses (HMGCR, NPC1L1, and PSCK9) also demonstrated that the combination of statins and ezetimibe, or their pharmacological targets, did not show a significant causal relationship with AD. Therefore, based on the current evidence, it can be concluded that there is no causal relationship between cholesterol-lowering drugs and AD.
Atherosclerosis and hyperlipidemia are recognized as significant risk factors for coronary heart disease and stroke. In the prevention and treatment of cardiovascular diseases, drug therapy, including lipid-lowering drugs, plays a crucial role (1). However, there have been studies suggesting that the use of lipid-lowering drugs may have an impact on cognitive function in elderly individuals, potentially increasing the risk of developing AD (2–4).
The prevalence of cognitive dysfunction and dementia, particularly AD, is on the rise in Europe due to the aging population (5–8). The pathophysiology of AD is complex and involves various mechanisms, including oxidative stress, inflammatory responses, and amyloid β-protein deposition (9–10). While aging and vascular-related diseases due to lipid accumulation are recognized as major risk factors for AD, there is evidence to suggest that the inappropriate use of statins may also interact with AD (11–12). In recent years, the incidence of AD has been steadily increasing, posing a significant health issue for elderly individuals and a societal challenge. Therefore, investigating the potential impact of lipid-lowering drugs on the development of AD is an important research direction (13–14).
Lipid-lowering drugs commonly used in clinical practice include statins, ezetimibe, and PCSK9 inhibitors, which are primarily used to lower low-density lipoprotein cholesterol (LDL-C) levels and reduce cardiovascular events. Some AD patients may also be prescribed statins for treatment (11, 13). However, long-term use of statins may lead to excessively low LDL-C levels, and there have been case reports indicating potential adverse effects on cognitive function, such as decreased reaction time and memory ability in elderly men, and an increased risk of depression (15–17). The US Food and Drug Administration has also added warning labels indicating possible cognitive impairment caused by statin drugs (6–7). Ezetimibe is a different type of cholesterol-lowering drug that inhibits cholesterol absorption in the intestines, and some evidence suggests that it may reduce β-amyloid deposition in the hippocampus area or reverse neurofibrillary tangles in AD rats (18–19). However, reports have also mentioned potential adverse psychiatric reactions, such as depression, memory loss, confusion, and aggressive behavior, associated with the use of lipid-lowering drugs (20). PCSK9 inhibitors are a newer class of lipid-lowering drugs that target PCSK9, and recent research indicates that they may interfere with brain cholesterol metabolism through synergistic effects with Aβ, while reducing neuronal cholesterol levels (21). Nevertheless, considering the mixed use of lipid-lowering drugs, along with differences in race, drug subtype, and study design, there is still considerable controversy over the safety of statin drugs (22–24).
In epidemiological research, cross-sectional and retrospective studies are commonly used methods. However, these studies may be affected by unmeasured confounding factors and bias, resulting in limited reliability and low statistical efficacy. Large-sample randomized controlled trials (RCTs) can control for some confounding variables but may have limitations in terms of evidence due to the long time it takes for AD pathogenesis to develop (25–26). Recently, there has been a growing interest in studying causality between exposure and disease using genome-wide association studies (GWAS) with the Mendelian randomization (MR) method. MR avoids bias from confounding factors, as alleles are randomly assigned. Additionally, genotypes often result in lifelong trait differences, making MR analysis useful for predicting the consequences of long-term exposure interventions (27–28).
The study utilized a two-sample MR method with single nucleotide polymorphism (SNP) data to investigate the causal relationship between statin therapy and AD, while avoiding inconsistent conclusions from previous studies.
Ethical consent was not required as the data used in the study were obtained from publicly available databases. Figure 1 provides an overview of the study design.
The GWAS data for this study were obtained from two databases: UK Biobank (UKB), Global Lipids Genetics Consortium (GLGC), and GWAS ATLAS resource. The Neale laboratory in UKB provided genetic data for different cholesterol-lowering therapies, including atorvastatin, pravastatin, rosuvastatin, simvastatin, and ezetimibe. The sample sizes for each therapy group were as follows: 13,851 cases and 449,082 controls for atorvastatin, 2,208 cases and 460,725 controls for pravastatin, 2,870 cases and 460,063 controls for rosuvastatin, 52,427 cases and 410,506 controls for simvastatin, and 1,997 cases and 335,162 controls for ezetimibe. Genetic data for LDL-C were obtained from the GLGC, which included 173,082 samples providing laboratory indicators. Data on AD were acquired from a previous meta-analysis study, which included a total of 71,880 AD cases and 383,378 control subjects. All genetic datasets used in this research involved individuals of European ethnicity.
In order to investigate the relationship between statin therapy and AD, we conducted a univariable MR analysis using SNPs associated with statin therapy. We limited our analysis to SNPs within 1Mb with r^2 < 0.001 aggregation. SNPs were considered valid instruments if they had a P-value less than 5×10^-8, except for pravastatin-associated SNPs which required a P-value less than 1×10^-5 due to limited availability of IV SNPs, and ezetimibe-associated SNPs which required a P-value less than 5×10^-9 due to a shrinkage threshold. For the multivariable MR analysis, the IV SNPs for ezetimibe and pravastatin were also limited with a P-value of 1×10^-5. We ensured that the Hardy-Weinberg Law was met for all SNPs (32), and details of all instrument variables (IV) SNPs used in the analysis are listed in Supplementary Table 1. To eliminate the potential confounding effects of other variables on AD through alternative pathways, we removed SNPs associated with recognized dysfunction, age, and neurological disease using the phenoscanner V2.0 database (33). Details of all GWAS resource conducted in our study are listed in Table 1.
Consortium |
Phenotype |
Population |
ncase |
ncontrol |
Sample size |
Number of SNPs |
---|---|---|---|---|---|---|
Neale Lab Consortium |
Atorvastatin therapy |
European |
13,851 |
449,082 |
462,933 |
9,851,867 |
Neale Lab Consortium |
Provastatin therapy |
European |
2,208 |
460,725 |
462,933 |
9,851,867 |
Neale Lab Consortium |
Rosuvastatin therapy |
European |
2,870 |
460,063 |
462,933 |
9,851,867 |
Neale Lab Consortium |
Simvastatin therapy |
European |
52,427 |
410,506 |
462,933 |
9,851,867 |
Neale Lab Consortium |
Ezetimibe therapy |
European |
1,997 |
335,162 |
337,159 |
10,894,596 |
Global Lipids Genetics Consortium |
LDL cholesterol |
Mixed |
173,082 |
/ |
173,082 |
2,437,752 |
A meta-analysis of GWAS |
Alzheimer Disease |
European |
71,880 |
383,378 |
455,258 |
13,367,301 |
Abbreviations: SNP, single nucleotide polymorphism; GWAS, genome-wide association studies.
For the univariable MR statistical analysis, we utilized various methods, including IVW (inverse-variance weighted), weighted median (WM), MR Egger, simple model, and weighted model, to investigate the causal relationship between GWAS effect alleles for statin therapy and AD. The IVW-MR method was primarily used to combine effect estimates by utilizing genetic variants associated with cholesterol lowering therapy as instruments. However, we also employed complementary methods such as MR Egger and WM to obtain more robust estimates in a wider range of scenarios, although these methods may have wider confidence intervals (CI). The "TwoSampleMR" package in R software version 4.2.2 was used for allele harmonization and analysis.
For the multivariable MR analysis, we conducted an extended analysis of the univariate MR, known as multivariable MR, which allows for joint testing of the causal effects of multiple risk factors (34). This was done to investigate the comprehensive effects of various lipid-lowering drugs on AD. Multivariable MR takes into consideration that patients taking lipid-lowering drugs often use them in combination, and that ezetimibe-related SNPs used in MR analysis may be related to statin-type lipid-lowering drugs. The SNPs used for multivariable MR are combinations of instrumental variables for each exposure, while confounding variables related to outcomes, cognition, or age were excluded. Software packages such as "TwoSampleMR", "MendelianRandomization", and "MVMR" in R were used for the MR analysis.
To identify potential genetic factors associated with LDL-C, we initially selected IV SNPs with a p-value less than 5×10^-8 and an r^2 value of less than 0.01. We then screened for candidate SNPs located within a ± 100 kb window of the targeted genes (HMGCR, NPC1L1, and PSCK9), ensuring that each SNP had a minor allele frequency greater than 1% (35). For single-SNP analyses, we used the Wald ratio method, while for multiple-SNP analyses, we employed instrumental variable weighted (IVW), MR-Egger, and WM methods to estimate effects using genetic variations linked to LDL-C levels as instruments. Heterogeneity and pleiotropy were assessed using established methods as described previously. Finally, allele coordination and analysis were performed using the "TwoSampleMR" package.
To assess the validity and robustness of the results, we performed sensitivity analyses. We used the MR-Egger intercept test to assess horizontal pleiotropy, which is the potential for a single genetic variant to influence multiple traits or outcomes. Heterogeneity was estimated by Cochran's Q test to determine if any single instrument drove the results and to check consistency with the MR hypothesis. Additionally, we used MR-PRESSO to detect outliers or causal direction bias between two samples when enough instrumental variables existed (36–37). All analyses were conducted using the "TwoSampleMR" package and "MR-PRESSO" package in R software version 4.2.2.
Overall, our results of MR estimates were not statistically significant, indicating that genetically predicted use of cholesterol-lowering therapy was not significantly associated with AD risk (Fig. 2).
In our analysis, we utilized two-sample MR methods to investigate the causal relationship between simvastatin, atorvastatin, pravastatin, rosuvastatin, and ezetimibe, and AD. We identified 32, 15, 6, 4, and 5 SNPs, respectively, associated with each treatment, excluding SNPs related to the outcome or confounding factors. However, our IVW analysis did not reveal any significant causal effects for simvastatin (OR = 1.138, 95% CI = 0.976–1.328, p = 0.098). MR Egger (OR = 1.189, 95% CI = 0.834–1.695, p = 0.347) and WM analysis (OR = 1.078, 95% CI = 0.867–1.339, p = 0.501) showed similar risk estimates. Sensitivity analyses using simple and weighted models did not yield significant changes. Additionally, no outliers or heterogeneity were detected in our study (MR Egger p = 0.260; IVW p = 0.299 from Cochran Q test; MR-PRESSO p = 0.109), and no directional pleiotropy was observed (intercept=-0.0003; SE = 0.001; p = 0.792). Similar analyses were conducted for other cholesterol therapies and AD, but no significant effects were found (IVW results: Atorvastatin: OR = 0.943, 95% CI = 0.612–1.453, p = 0.789; Pravastatin: OR = 6.857, 95% CI = 0.514–90.864, p = 0.144; Rosuvastatin: OR = 2.466, 95% CI = 0.333–18.278, p = 0.377; Ezetimibe: OR = 1.292, 95% CI = 0.239-6,969, p = 0.766). Detailed data of IVW, MR Egger, and WM are presented in Table 2. Our results did not reveal any horizontal pleiotropy, outliers, or heterogeneity (Table 3).
Exposure |
Method |
SNPs |
OR |
Low CI of OR |
Up CI of OR |
P_value |
---|---|---|---|---|---|---|
Atorvastatin therapy |
MR Egger |
15 |
0.758 |
0.217 |
2.654 |
0.672 |
Atorvastatin therapy |
Weighted median |
15 |
0.870 |
0.478 |
1.582 |
0.641 |
Atorvastatin therapy |
Inverse variance weighted |
15 |
0.943 |
0.612 |
1.453 |
0.789 |
Atorvastatin therapy |
Simple mode |
15 |
0.893 |
0.34 |
2.342 |
0.821 |
Atorvastatin therapy |
Weighted mode |
15 |
0.856 |
0.389 |
1.883 |
0.705 |
Provastatin therapy |
MR Egger |
6 |
230054.811 |
<0.001 |
8.91753E + 14 |
0.335 |
Provastatin therapy |
Weighted median |
6 |
6.188 |
0.234 |
163.821 |
0.276 |
Provastatin therapy |
Inverse variance weighted |
6 |
6.857 |
0.517 |
90.864 |
0.144 |
Provastatin therapy |
Simple mode |
6 |
4.964 |
0.035 |
704.471 |
0.554 |
Provastatin therapy |
Weighted mode |
6 |
4.272 |
0.033 |
554.371 |
0.584 |
Rosuvastatin therapy |
MR Egger |
4 |
4.794 |
<0.001 |
3045917252 |
0.893 |
Rosuvastatin therapy |
Weighted median |
4 |
2.582 |
0.252 |
26.413 |
0.424 |
Rosuvastatin therapy |
Inverse variance weighted |
4 |
2.466 |
0.333 |
18.278 |
0.377 |
Rosuvastatin therapy |
Simple mode |
4 |
3.128 |
0.126 |
77.628 |
0.537 |
Rosuvastatin therapy |
Weighted mode |
4 |
2.925 |
0.115 |
74.161 |
0.562 |
Simvastatin therapy |
MR Egger |
32 |
1.189 |
0.834 |
1.695 |
0.347 |
Simvastatin therapy |
Weighted median |
32 |
1.078 |
0.867 |
1.339 |
0.501 |
Simvastatin therapy |
Inverse variance weighted |
32 |
1.138 |
0.976 |
1.328 |
0.098 |
Simvastatin therapy |
Simple mode |
32 |
1.132 |
0.749 |
1.710 |
0.562 |
Simvastatin therapy |
Weighted mode |
32 |
1.120 |
0.846 |
1.481 |
0.435 |
Ezetimibe therapy |
MR Egger |
4 |
2.835 |
0.025 |
317.27 |
0.707 |
Ezetimibe therapy |
Weighted median |
4 |
1.055 |
0.152 |
7.336 |
0.957 |
Ezetimibe therapy |
Inverse variance weighted |
4 |
1.292 |
0.239 |
6.969 |
0.766 |
Ezetimibe therapy |
Simple mode |
4 |
0.857 |
0.061 |
12.052 |
0.916 |
Ezetimibe therapy |
Weighted mode |
4 |
0.810 |
0.064 |
10.257 |
0.881 |
Abbreviations: SNP, single nucleotide polymorphism; MR, mendelian randomization; OR, odds ratio; CI, confidence interval. |
We conducted a comprehensive multivariable MR study to investigate the effects of different statins and ezetimibe use on AD, taking into consideration that many patients are now using combination therapy. Firstly, we screened for significant exposure variables in multiple GWAS databases and controlled for confounding factors using the phenoscanner V2.0 database. Next, we analyzed the causal effect of statins and ezetimibe on AD. The results from inverse variance weighted (IVW) analysis showed no significant difference in the incidence of AD among different subtypes of statins when combined with ezetimibe (Simvastatin: OR = 0.758, 95% CI = 0.217–2.654, p = 0.461; Atorvastatin: OR = 0.758, 95% CI = 0.217–2.654, p = 0.782; Provastatin: OR = 0.758, 95% CI = 0.217–2.654, p = 0.689; Rosuvastatin: OR = 0.758, 95% CI = 0.217–2.654, p = 0.592). Furthermore, our multivariable MR analysis indicated that ezetimibe also did not affect the incidence of AD. Both Multivariable WM and Egger analyses revealed similar results. Additionally, MR-PRESSO and Cochran Q-test tests demonstrated no multicollinearity or heterogeneity, except in the Simvastatin & Ezetimibe group in our Multivariable MR analyses (Table 4).
Exposure |
N.SNPs |
MVMR-IVW OR (95% CI) |
P_value |
MVMR-Egger OR (95% CI) |
P_value |
MVMR-WM OR (95% CI) |
P value |
Egger (intercept) value |
Cochran Q P_value |
---|---|---|---|---|---|---|---|---|---|
Simvastatin & Ezetimibe therapy |
28 |
0.334 |
0.002 |
||||||
Simvastatin |
1.196(0.742, 1.927) |
0.461 |
1.513(0.734, 3.725) |
0.225 |
1.704(0.975, 2.977) |
0.061 |
|||
Ezetimibe |
0.318(0.007, 15.502) |
0.564 |
8.551(0.002, 9.728) |
0.368 |
0.034(0, 2.956) |
0.138 |
|||
Atorvastatin & Ezetimibe therapy |
15 |
0.186 |
0.443 |
||||||
Atorvastatin |
1.183(0.359, 3.904) |
0.782 |
3.206(0.48, 21.413) |
0.229 |
1.084(0.226, 5.202) |
0.919 |
|||
Ezetimibe |
0.931(0.008, 105.425) |
0.976 |
0.147(0.001, 34.674) |
0.492 |
0.57(0.001, 268.54) |
0.858 |
|||
Provastatin & Ezetimibe therapy |
22 |
0.508 |
0.966 |
||||||
Provastatin |
1.298(0.361, 4.674) |
0.689 |
2.678(0.22, 32.525) |
0.440 |
2.337(0.446, 12.244) |
0.315 |
|||
Ezetimibe |
8.534(0.785, 92.759) |
0.078 |
10.444(0.892, 122.119) |
0.062 |
5.618(0.23, 137.277) |
0.290 |
|||
Rosuvastatin & Ezetimibe therapy |
4 |
1.000 |
0.334 |
||||||
Rosuvastatin |
0.055(0, 2188.562) |
0.592 |
0.055(0, 3158.966) |
0.605 |
0.055(0, 2188.562) |
0.885 |
|||
Ezetimibe |
36.162(0.002, 815861.59) |
0.483 |
36.27(0, 1.56e + 11) |
0.751 |
36.162(0.002, 815861.59) |
0.793 |
|||
Abbreviations: SNP, single nucleotide polymorphism; MVMR, multivariable mendelian randomization; OR, odds ratio; CI, confidence interval; IVW, inverse variance weighted; N, number; WM, Weighted median. |
After screening LDL-related cholesterol-lowering targets, our results did not provide suggestive evidence for a link between these targets (HMGCR, NPC1L1, and PCSK9) mediated LDL cholesterol (equivalent to a 1 mmol/L increase) and the incidence of AD. For HMGCR and NPC1L1, only one SNP remained after Wald ratio analysis was conducted (HMGCR: OR = 0.998, 95% CI = 0.942–1.059, p = 0.960; NPC1L1: OR = 1.019, 95% CI = 0.914–1.136, p = 0.736). For PCSK9, three SNPs were finally retained. The IVW-MR analysis also revealed no causal relationship (OR = 0.995, 95% CI = 0.965–1.026, p = 0.737) (Table 5). Additionally, WM and MR Egger analyses failed to provide any evidence suggesting that targeting lipid-lowering drugs could potentially impact AD. This finding further supports the idea that lipid-lowering therapy may not cause AD through its pharmacological targets.
Exposure |
Method |
Causal Estimate (SD) |
Intercept (SE) |
P_value |
---|---|---|---|---|
Atorvastatin therapy |
MR-PRESSO |
0.059(0.172) |
0.736 |
|
Atorvastatin therapy |
MR Egger |
0.001(0.002) |
0.723 |
|
Atorvastatin therapy |
Cochran Q MR Egger |
0.819 |
||
Atorvastatin therapy |
Cochran Q IVW |
0.862 |
||
Provastatin therapy |
MR-PRESSO |
1.925(0.983) |
0.107 |
|
Provastatin therapy |
MR Egger |
-0.008(0.009) |
0.404 |
|
Provastatin therapy |
Cochran Q MR Egger |
0.752 |
||
Provastatin therapy |
Cochran Q IVW |
0.734 |
||
Rosuvastatin therapy |
MR-PRESSO |
0.903(0.684) |
0.278 |
|
Rosuvastatin therapy |
MR Egger |
-0.001(0.013) |
0.954 |
|
Rosuvastatin therapy |
Cochran Q MR Egger |
0.512 |
||
Rosuvastatin therapy |
Cochran Q IVW |
0.719 |
||
Simvastatin therapy |
MR-PRESSO |
0.130(0.078) |
0.109 |
|
Simvastatin therapy |
MR Egger |
-0.0003(0.001) |
0.792 |
|
Simvastatin therapy |
Cochran Q MR Egger |
0.260 |
||
Simvastatin therapy |
Cochran Q IVW |
0.299 |
||
Ezetimibe therapy |
MR-PRESSO |
0.256(0.598) |
0.698 |
|
Ezetimibe therapy |
MR Egger |
-0.002(0.004) |
0.760 |
|
Ezetimibe therapy |
Cochran Q MR Egger |
0.515 |
||
Ezetimibe therapy |
Cochran Q IVW |
0.694 |
Abbreviations: SNP, single nucleotide polymorphism; MR, mendelian randomization; OR, odds ratio; CI, confidence interval; IVW, inverse variance weighted; LDL-C, low-density lipoprotein cholesterol; HMGCR, 3-Hydroxy-3-Methylglutaryl Coenzyme A Reductase; NPC1L1, Niemann-Pick disease, type C1; PCSK9, Proprotein Convertase Subtilisin/Kexin Type 9.
Cardiovascular diseases are becoming increasingly prevalent among the elderly, with 17.9 million deaths attributed to them in 2016, accounting for 31% of global deaths, as reported by the World Health Organization (WHO) (1). The incidence of cardiovascular disease is rising worldwide and poses a significant economic and social burden. Patients with cardiovascular and cerebrovascular diseases often require long-term medication therapy to lower their LDL-C cholesterol levels (38–39). Statins are recommended as first-line agents for adults with LDL-C levels greater than 190 mg/dl and diabetics with levels ranging from 70–190 mg/dl, as they have a low risk of adverse events and are effective in primary prevention of cardiovascular disease (40–42).
The standardized use of lipid-lowering drugs remains a topic of debate due to potential bias from study design in epidemiological studies, which can only establish correlation without elucidating temporal order or causality (27–28). In this study, the use of MR methods revealed causal relationships more accurately, without considering potential bias from study design. Our conclusion is consistent with previous high-quality meta-analyses. Using two sets of genetic instruments and excluding confounding factors, genetic evidence indicated no causal effect of statins on AD (AD) development (36–37). Furthermore, we found no significant causal relationship between statin or ezetimibe alone and the incidence of AD. Our results remained unchanged when considering the combination of cholesterol-lowering therapy. Carlos et al. demonstrated in a narrative review that most of the evidence for cognitive impairment caused by lipid-lowering drugs comes almost entirely from case reports (43). Although these case reports help identify potential clinical drug risks, they are limited by their observational nature, and it is important to interpret these reports accordingly (43). It is worth noting that the benefits and risks of statin use have been well established in several large-scale clinical cohort studies, showing a significant reduction in cardiovascular events, especially when used for secondary prevention (44–45). Currently, no large-sample clinical studies have found that statins can further impair existing cognitive impairment. Data from animal models and limited human data suggest that statin use may have cognitive benefits. The specific mechanism may be related to lowering plasma cholesterol levels, inhibiting cerebral vascular plaque formation, as well as non-cholesterol-related pathways such as alleviating endothelial dysfunction, increasing endothelial nitric oxide generation, anti-inflammatory effects, and antioxidant effects (23.25).
Regarding ezetimibe therapy, there are currently no reports of potential cognitive impairment effects. Existing research suggests that it may prevent further deterioration of cognition in rats with hyperlipidemia or AD rats, but more evidence is needed to determine its specific impact on cognition (18–19). PCSK9 inhibitors are the latest class of cholesterol-lowering drugs. A study conducted on a new monoclonal antibody against PCSK9 (LY3015014) showed that it did not have a significant impact on cognitive function when compared to placebo controls. However, the study was short-term in nature. Another ongoing prospective study included over 2,000 participants who were taking statins and PCSK9 inhibitor (evolocumab). However, no significant differences in cognition were observed during a series of cognitive tests, even when LDL-C levels reached the lowest point (below 0.65 mmol/L) for some participants (46–47). Olmastroni et al.'s systematic review also suggested that the benefit-harm balance of cholesterol-lowering drugs is generally favorable (42). In addition, lifestyle factors such as obesity, long-term smoking, alcohol abuse, and educational attainment may play more important roles in triggering AD compared to cholesterol-lowering therapy alone, but strong clinical evidence is still lacking (10–13).
However, there are still some key factors to consider when explaining the association between lipid-lowering drugs and AD. The first factor is the clinical significance of study results. For example, in one study, users of statins had a corrected Mini-Mental State Examination (MMSE) score of 93.7 compared to non-users who had a score of 92.7. Although these results were statistically significant, the difference was clinically irrelevant (48). The second factor to consider is the quality of studies involved. Multiple systematic reviews have shown that studies involving the effects of statins on cognitive impairment or AD are mostly observational or retrospective studies, and randomized controlled trials and stronger evidence are needed to determine their potential risks or benefits (41–42). Thirdly, statins or other lipid-lowering targeted drugs have not yet been explicitly approved by the FDA or Health Canada for the prevention or treatment of cognitive impairment.
In the past, LDL-C was considered a biomarker for cognitive impairment (49); however, current evidence suggests that normal doses of statins alone do not lead to very low LDL-C levels in the central nervous system due to the protection of the blood-brain barrier, which is maintained by cholesterol-like hormones and bile acids produced by the body (50–51). We also conducted an analysis on whether different lipid-lowering drugs that act on LDL-C causally induce AD. Interestingly, our results showed that the odds ratio (OR) values were very close to 1. Based on MR associations of both the drug itself and the drug targets with AD, neither of them were found to be significant, further confirming that the pharmacological effect of cholesterol-lowering therapy is unlikely to induce or aggravate the development of AD. It is worth noting that previous animal experiments have even found that users of PCSK9 inhibitors may have an increased risk of developing AD, suggesting that PCSK9 dysfunction may result in brain β-amyloid protein production and neuronal cell death, serving as a biomarker for AD (26). Therefore, in addition to reliable clinical evaluations of traditional statin drugs, we should not ignore the interaction between new lipid-lowering drugs and AD, as well as the potential cognitive impairment they may induce. Finally, it should be noted that although cognitive impairment is a rare side effect of lipid-lowering drugs, the clinical use by a large number of patients can amplify rare adverse reactions hundreds or thousands of times over. Therefore, standardized use of lipid-lowering drugs requires long-term follow-up investigations, observational trials, and randomized controlled trials with specific designs.
Our study has several limitations. Firstly, we were unable to directly analyze the causal relationship between the use of PCSK9 inhibitors and the incidence of AD due to the lack of related GWAS data. Secondly, MR analysis is limited by the population from which genetic data is obtained; therefore, our findings may not be representative of the entire British or European population, and there may be racial genetic differences between populations. Thirdly, some subgroups of SNPs for statin drugs have small sample sizes, which may not provide accurate correlations. Finally, since different types of statin drugs have varying pharmacological mechanisms and limited ability to cross the blood-brain barrier, our findings may not capture all tissue-specific relationships for all types of statin drugs.
Using MR analysis to investigate various types of cholesterol-lowering therapies and their association with AD endpoints, our study found no causal relationship between them. This suggests that neurological effects caused by lipid-lowering medications are rare occurrences that can be managed within the overall population. However, it is still crucial for clinical doctors to accurately assess the indications when prescribing lipid-lowering drugs for elderly patients, and to make appropriate selections of drugs based on past evidence-based medicine in order to minimize the risk of rare cognitive impairment.
Acknowledgements
Not applicable.
Disclosure of interest
The authors declared that there were no conflict of interest.
Funding
This work was supported in part by the Natural Science Foundation of Nanjing University of Traditional Chinese Medicine (grant no. XZR2020072) and Nanjing Health Science and Technology Development Special Fund Project (YKK21118).
Ethics approval
Not applicable.
Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.
Authors' contributions
SH and GY contributed to the conception of the study. GY contributed to the funding acquisition. SH, JG and GY performed the analysis. SH and JG contributed to data collection and screening. SH, JG contributed to the design of methodology and use of software. SH and GY provided resources, supervised the study and wrote the original draft. GY confirm the authenticity of all the raw data. All authors read and approved the final manuscript.