Introduction

Osteoarthritis(OA) is a degenerative joint disease related to cartilage that affects approximately 300 million people worldwide [1]. It predominantly affects the knee, hand, and hip [2], resulting in chronic pain, joint stiffness, and mobility disorders that significantly affect patient’s quality of life. Various risk factors contribute to OA, including age, smoking, body mass index(BMI), low-density lipoprotein(LDL), and alcohol consumption, previous studies have also revealed the role of hypertension in the development of knee OA [3,4,5]. Therefore, controlling blood pressure may be a crucial factor in preventing OA.

Previous studies suggest that the homeostasis regulation of chondrocytes depends on the local renin–angiotensin system. Thus, angiotensin-converting enzyme (ACE) inhibitors or angiotensin-II receptor blockers may have therapeutic potential for knee OA [6]. Li et al. found that the use of calcium channel blockers is associated with a narrowed knee joint space [7], while Driban et al. observed that thiazides are correlated with changes in symptoms in patients with knee OA [8]. Although various studies have indicated the value of antihypertensive drugs in the prevention and treatment of OA, sample size and experimental design limitations have made it challenging to establish causal relationships between drug use and changes in OA-related traits. Moreover, observational studies are susceptible to mixed factors and reverse causal relationships, leading to biased results. Mendelian randomization(MR) provides a novel approach to address these issues.

Mendelian randomization is a statistical method that uses genetic variation as an instrumental variable to estimate the causal relationship between exposure and outcome [9]. Because genetic variation is randomly distributed during conception, it is unlikely to be influenced by environmental and confounding factors, thereby minimizing confounding bias and providing robust evidence of causal relationships. Drug-targeted Mendelian randomization is a research method that uses drug target-related genetic variation to simulate drug effects. Several studies have successfully evaluated the relationship between antihypertensive drugs and Alzheimer's disease [10, 11] and other diseases [12, 13]. In this study, we used data from published studies as instrumental variables to comprehensively evaluate the relationship between antihypertensive drugs and OA and assess their potential therapeutic value.

Methods and design

Research design

We conducted a two-sample Mendelian randomized method. Initially, we assessed the causal relationship between systolic blood pressure(SBP) and diastolic blood pressure(DBP) and all OA, hand OA, hip OA, and knee OA. Next, we employed the multivariate Mendelian randomized(MVMR) study method to examine the impact of blood pressure on these four types of OA after multivariate adjustment. Finally, we utilized published genetic variation of antihypertensive drug targets as instrumental variables to evaluate the effects of 12 types of antihypertensive drugs on OA. These drugs included alpha-adrenoceptor blockers, angiotensin-converting enzyme inhibitors, angiotensin-II receptor blockers, beta-adrenoceptor blockers, calcium channel blockers (CCB), centrally acting antihypertensive drugs, loop diuretics, potassium-sparing diuretics and aldosterone antagonists, renin inhibitors, thiazides and related diuretics, and vasodilator antihypertensives. We also used the independent cohort of osteoarthritis in the UK Biobank to conduct a validation analysis.

Selection of data sources and instrumental variables

Systolic and diastolic blood pressure

We obtained systolic and diastolic blood pressure phenotypes from the IEU Open GWAS (DBP: ukb-a-359, SBP: ukb-a-360), which included 317,756 samples from European populations. We selected single nucleotide polymorphisms(SNPs) closely related to DBP and SBP (p < 5 × 10–8) and removed those in linkage disequilibrium (r2 = 0.001, genetic distance = 10 MB). We then eliminated SNPs related to OA, BMI, smoking, obesity, drinking, and low density lipoprotein (LDL) through the Phenoscanner data website (http://www.phenoscanner.medschl.cam.ac.uk/) [14]. We calculated the F value of each remaining SNP and removed those with an F value less than 10. Finally, we used the MR-PRESSO method to screen and eliminate outliers and used the remaining SNPs as the final instrumental variables.

Antihypertensive drugs

We selected 12 genetic proxies for antihypertensive drugs from previously published studies [15]. In this research, authors initially extracted protein targets and genes of 12 antihypertensive drugs from the Drugbank database, following which they screened the corresponding gene's "best SNP" in each tissue using the GTeX project data. After conducting two-sample MR and performing column quality control, the researchers selected SNPs that significantly affected SBP as the final analysis SNP. The detailed process of SNP screening is available in the original study by the authors [15]. To circumvent confounding factors, we scrutinized and eliminated SNPs linked to OA, BMI, obesity, smoking, drinking, and LDL in the Phenoscanner. We further employed the MR-PRESSO method to screen and discard any outliers.

Outcome

We extracted OA GWAS data from the Genetics of Osteoarthritis (GO) Consortium [16], which included 177,517 OA cases. Hand OA, hip OA, and knee OA were 20,901 cases, 36,445 cases, and 62,497 cases, respectively, all of which were European population samples. The GWAS data of the validation cohort is from the UK Biobank. Due to the lack of GWAS data of hand Osteoarthritis, we only conducted validation analysis on all OA, knee OA and hip OA in the outcome.

Statistical analysis

We utilized three different statistical methods: inverse variance weighting(IVW), weighted median estimator (WME), and MR Egger regression. Specifically, IVW was the primary analysis method, while the other two methods supplemented our analysis. Heterogeneity was detected using Cochran's Q test. We evaluated the bias of gene pleiotropy using MR Egger intercept analysis. If the regression intercept is closer to 0, gene pleiotropy is less likely to have occurred. We also used a sensitivity analysis known as leave-one-out. We replicated the above method for the verification cohort. To ensure independent comparison based on MR, we established a machine learning model using the harmonized data. For the analysis, we employed the linear regression model in this test set. To examine the impact of effect size on the outcome of exposure factors, we randomly split the data into a training set (80%) and a test set (20%). We considered a p value less than 0.05 as statistically significant.

All analyses were conducted using RStudio 2023.03.1, R 4.22, TwoSample MR, and the MR PRESSO package. In univariate MR analysis of blood pressure and OA, we considered a p value of < 0.006 (0.05/8) to be significantly different after Bonferroni correction. In MVMR, we considered a p value of < 0.05 to be significantly different. In antihypertensive drugs and OA analysis, a p value of < 0.001(0.05/48) was considered to be significantly different. Any p value between 0.05 and significant results was considered to be nominally significant.

Results

Instrumental variables

Following screening, we employed 168, 157, and 289 SNPs, respectively, as instrumental variables for DBP, SBP, and antihypertensive drugs. Further information regarding the instrumental variables can be accessed in the supplemental file.

MR analysis of blood pressure and OA

Figure 1 displays the results of the univariate MR analysis. Our primary IVW analysis did not identify evidence that blood pressure substantially influences OA. However, we observed nominally significant evidence of a relationship between SBP and hand OA, with an OR of 0.852 (95% CI 0.745–0.974, P = 0.0193). In the MVMR analysis in Fig. 2, we did not observe evidence that blood pressure influences OA.

Fig. 1
figure 1

Univariate MR analysis results of blood pressure and OA

Full size image
Fig. 2
figure 2

MR analysis results of the relationship between blood pressure and OA after multivariable adjustment

Full size image

Antihypertensive drugs and OA

In our analysis of the relationship between antihypertensive drugs and OA, we found significant evidence supporting a causal association between PSDs and aldosterone antagonists with all OA (OR: 0.560, 95% CI 0.406–0.772, P = 0.0004), as well as hand OA (OR: 0.125, 95% CI 0.045–0.344, P = 0.0001) and knee OA (OR: 0.260, 95% CI 0.155–0.434, P = 0.0000) in our primary analysis. Consistent and significant or suggestive results were also observed in the other two methods. Specifically, we also observed a significant relationship between adrenergic neurone blockers and knee OA in our primary analysis (OR: 0.280, 95% CI 0.131–0.599, P = 0.001) (Fig. 3).

Fig. 3
figure 3

MR analysis results of the relationship between antihypertensive drugs and OA

Full size image

In our sensitivity analysis, we observed horizontal pleiotropic effects during the analysis of PSD and aldosterone antagonists for both hand and knee OA, as well as during the analysis of Adrenergic neurone blockers for knee OA. However, we found no such effects or heterogeneity during the MR analysis of PSDs and aldosterone antagonists with all OA (Fig. 3). In addition, our leave-one-out analysis further supports the reliability of our results (Fig. 4).

Fig. 4
figure 4

MR leave−one−out sensitivity analysis for PSDs and aldosterone antagonists on AllOA

Full size image

Cohort validation

No evidence of a causal relationship between blood pressure and osteoarthritis was observed in the validation cohort. However, extensive heterogeneity was found in the sensitivity analysis (Table 5 of the supplementary). Among antihypertensive drugs, we identified a significant causal relationship between calcium channel blockers and all OA (OR: 1.038, 95% CI 1.019–1.059, P = 0.0001). We also found a significant causal relationship between PSDs and knee OA (OR: 0.184, 95% CI 0.089–0.378, P = 0.0000). However, horizontal pleiotropy was observed in the MR analysis of PSDs and aldosterone antagonists with knee OA (P = 0.013). Furthermore, a nominally significant causal relationship was found between PSDs and aldosterone antagonists with all OA (OR: 0.935, 95% CI 0.892–0.980, P = 0.0055) and hip OA (OR: 2.244, 95% CI 1.008–4.995, P = 0.0478). The MR analysis of PSDs and aldosterone antagonists with hip OA also revealed horizontal pleiotropy (P = 0.013). In addition, a nominally significant causal relationship was found between loop diuretics and knee OA (OR: 0.149, 95% CI 0.035–0.643, P = 0.0107). The results of the validation cohort are presented in Table 6 of the supplementary.

Linear regression results

The linear regression results provide evidence for the protective effect of PSDs and aldosterone antagonists on all OA (β = −0.431, P = 0.000). Furthermore, the regression results also indicate a protective effect of PSDs and aldosterone antagonists on hand OA and knee OA. Moreover, our results demonstrated the influence of antihypertensive drugs, such as calcium channel blockers and beta adrenoceptor blockers, on OA. Detailed linear regression results are provided in Table 7 of supplementary.

Discussion

Our MR study did not yield any evidence to suggest a significant association between blood pressure and OA. However, we identified a causal relationship between PSDs and aldosterone antagonists in relation to all OA, indicating a protective effect against OA. This result was nominally replicated with statistical significance in the validation data set and exhibited significant correlation in the linear regression analysis. Furthermore, our sensitivity analysis provides additional support for the reliability of these findings.

The relationship between blood pressure and OA has been extensively studied. A previous study found a gender difference in the association between hypertension and the imaging of OA [3]. In addition, a Mendelian randomization study showed that SBP was associated with all OA, hip OA and knee OA [17]. However, a cross-sectional study conducted in South Korea showed that hypertension was negatively correlated with the prevalence of OA [18], which may be related to the continuous use of antihypertensive drugs. Wang et al.'s research [19] revealed a correlation between blood pressure and arterial stiffness at various locations, as well as the volume of knee cartilage, suggesting that blood pressure may affect the progression of OA. Despite several studies indicating a relationship between blood pressure and OA, our analysis did not find significant evidence supporting the idea that blood pressure affects OA. However, specific antihypertensive drugs were found to have a protective effect on OA, which suggests that they may play a role in protecting against OA by means other than simply reducing blood pressure.

In our study, PSDs and aldosterone antagonists were thoroughly validated and found to have a causal relationship with OA. However, previous studies have also indicated a potential association between other antihypertensive drugs and OA. A study using data from the Osteoarthritis initiative demonstrated that CCB may exacerbate pain in patients with OA. The findings indicate that, during the 4-year follow-up period, the pain score of the CCB group was higher compared to other antihypertensive drugs. Although no statistical difference was observed, the CCB group exhibited the narrowest joint gap across multiple test groups [7]. In another study [20], it was found that Nifedipine, a classic CCB, may have a detrimental effect on chondrocyte proliferation. In addition, previous studies [7, 21, 22] have suggested that beta-adrenoceptor blockers can alleviate joint pain. Furthermore, a systemic review [23] on losartan and Angiotensin-II receptor blockers indicates that Angiotensin-II receptor blockers have beneficial effects on chondrocytes in animals, suggesting a potential protective effect against Osteoarthritis.

OA is considered to be closely associated with a series of inflammatory reactions. Prior research has demonstrated that Aldosterone plays a role in promoting inflammatory effects [24]. Consequently, blocking Aldosterone may be beneficial in ameliorating the pain and other symptoms experienced by OA patients. Elsaman et al.’s investigation indicated that knee joint effusion related to OA can be improved using low-dose Spironolactone [25]. A previous randomized controlled trial (RCT) [26] also indicated significant improvement in quality of life among subjects who orally took 25 mg of Spironolactone every day. This improvement was demonstrated 5 month post-treatment, with about one-third of the Spironolactone group reporting pain and discomfort relief. Conversely, another proof-of-concept test involving 86 subjects based on this study found that taking Spironolactone does not significantly improve pain or quality of life among knee OA patients, whose average age surpassed 77 years old [27]. However, the study did indicate that taking 25 mg Spironolactone daily is highly safe for OA patients in this age group. Although our research demonstrates the potential efficacy of Aldosterone antagonists in treating OA, high-quality RCT verification is still necessary.

Our study's advantage is utilizing MR to investigate the effects of blood pressure and antihypertensive drugs on OA. By upholding MR's basic hypotheses, we minimized bias and confounding factors to a great extent. In addition, we employed publicly available data in conjunction with our research design to screen out instrumental variables, ensuring Instrumental Variables Estimation's maximum reliability. We further validated the results’ dependability via sensitivity analysis.

However, this research also has its limitations: first, the GWAS data we used was limited to individuals of European ancestry. Unfortunately, we could not explore the effects of blood pressure and antihypertensive drugs on other races, since sufficient data were not available. Second, the ability of Instrumental Variables Estimation to explain related genetic traits is limited. Finally, as MR studies primarily deal with assessing associated effects, additional high-quality research is needed to verify the improvement of antihypertensive drugs on OA.

Conclusion

The genetic prediction of PSDs and aldosterone antagonists suggested their protective effects against OA, providing essential guidance for the redevelopment of current drugs and therapy for hypertension in osteoarthritic patients. Nonetheless, additional high-quality research is still necessary for confirmation.