Osteoarthritis and hypertension: observational and Mendelian randomization analyses

NHANES constitutes an ongoing series of cross-sectional surveys aimed at evaluating the health and nutritional status of the civilians in the United States. This initiative is overseen by the National Center for Health Statistics under the purview of the Centers for Disease Control and Prevention (CDC). Additionally, it has received the ethical approval of the Ethics Review Board at the National Center for Health Statistics [15]. All procedures were performed following the guidelines of the Declaration of Helsinki.

In order to investigate the potential connection between osteoarthritis and hypertension, we conducted a retrospective analysis using data gathered from participants (≥ 20 years of age) across four 2-year NHANES survey cycles (2013–2020). The proportion of missing values for all variables was less than 15%, and missing data were imputed using the package “missforest” in R.

Osteoarthritis was denoted as M19.9 based on the International Classification of Diseases 10th Revision (ICD-10). Hypertension was defined as the average of three systolic blood pressure measurements ≥ 140 mmHg, the average of three diastolic blood pressure measurements ≥ 90 mmHg, current usage of medications for blood pressure control, or previously told had high blood pressure [16].

The following covariates were included in this study: age, gender, NHANES cycles, race, educational level, family income, heavy drinker, smoker, BMI, uric acid, high-density lipoprotein cholesterol (HDL-C), total cholesterol (TC), triglyceride (TG), blood urea nitrogen (BUN), aspartate transaminase (AST), alanine aminotransferase (ALT), glycohemoglobin, estimate glomerular filtration rate (eGFR), cancer, diabetes mellitus (DM), osteoporosis, stroke, arteriosclerotic cardiovascular disease (ASCVD), heart attack, CHF, walk or bicycle, work activity, recreational activity, antiplatelet agents, statins, antihypertensive agents. Family income was divided into three levels according to the family poverty income ratio: low income (≤ 1.3), medium income (> 1.3 to 3.5), and high income (> 3.5) [17]. The definition of smoker, heavy drinker, ASCVD, and DM as described in previous literature [18]. Data on diagnosed history of cancer, osteoporosis, stroke, heart attack, and CHF were self-reported. Physical activity including walk or bicycle, work activity, and recreational activity was obtained in questionnaire. Specifically, work activity and recreational activity were classified into four categories based on NHANES guidelines: none, moderate, both, and vigorous. The use of antiplatelet agents, statins, and antihypertensive agents was derived from medication history.

According to NHANES analytic guidelines, complex sampling design and sampling weights were considered in this study [19]. T tests and Chi-square tests were applied for analyzing the association of hypertension with continuous variables and categorical variables, respectively. The multivariable logistic regression was employed to compute odds ratios (ORs) and 95% confidence intervals (95% CIs) to ascertain the effects of osteoarthritis on hypertension. In the logistic regression model, Model 1 was adjusted for none. Model 2 was adjusted for age, gender, heavy drinker, smoker, and education level. Model 3 was further adjusted for BMI, HDL-C, TC, BUN, ALT, glycohemoglobin, eGFR, family income, cancer, ASCVD, DM, osteoporosis, stroke, heart attack, CHF, uric acid, antiplatelet agents, and statins. Model 4 was based on model 3 and walk or bicycle. Confounding factors that were significantly associated with hypertension (P < 0.05) or a change in effect estimate of more than 10% were included in Model 4 [20]. The variance inflation factor (VIF) for all variables in the models was below 10, mitigating the impact of multicollinearity on the results [21]. Subgroup analyses were performed to evaluate the association between osteoarthritis and hypertension across various subgroups. And likelihood ratio tests were employed to identify potential interactions. Subsequently, propensity score matching (PSM) was conducted, the matching variables encompassed age, gender, heavy drinker, smoker, Race, education level, BMI, NHANES cycles, and walk or bicycle with caliper values set at 0.02. Final, logistic regression analysis was performed on the data after PSM. Model 1was adjusted for none. Model 2 was adjusted for age, gender, BMI, HDL-C, TG, BUN, ALT, AST, glycohemoglobin, eGFR, family income, uric acid, ASCVD, DM, stroke, heart attack, CHF, antiplatelet agents, recreational activity and statins. All analyses were performed using R 4.1.2 (http://​www.​r-project.​org). Two-sided levels of significance were calculated, and the significance level was set as 0.05.

To ensure the accuracy and robustness of causal inference, the selection of genetic instruments should satisfy the three key assumptions of MR. First, Single Nucleotide Polymorphisms (SNPs) were chosen based on a genome-wide significance threshold and F-statistics > 10 [22]. Second, independent SNPs were kept based on linkage disequilibrium (LD) as measured by r² < 0.001 and clumping distance = 10,000 kb. Final, SNPs associating with the outcome and confounding factors (p < 5 × 10^− 8) were excluded, and palindromic SNPs with intermediate allele frequencies were deleted when harmonizing exposure and outcome data [23, 24].

The OA data were sourced from the UK Biobank (UKB) [25], which serves as a crucial biobank resource (http://​www.​nealelab.​is/​uk-biobank/​) and from the GWAS data analyzed by Tachmazidou et al. [26]. In the dataset obtained from Tachmazidou, three distinct phenotypes were employed in our study: knee osteoarthritis, hip osteoarthritis, and knee or hip osteoarthritis, as detailed in Table S1. When osteoarthritis was considered as the exposure variable, the significance threshold was set at p < 5 × 10^− 7. SNPs associated with hypertension or with confounding factors such as body weight, BMI, and obesity were systematically excluded through the use of PhenoScanner [27].

The sources of GWAS data for hypertension encompassed the ninth release of the FinnGen Study [28]. In the FinnGen Study, the hypertension dataset consisted of 111,581 cases and 265,626 controls. When hypertension was considered as the exposure, we employed a more stringent significance threshold (p < 1 × 10^− 8) due to the numerous SNPs associated with hypertension. Furthermore, SNPs associated with osteoarthritis or with confounding factors (bone density, body weight, BMI, obesity, and diabetes) were excluded using PhenoScanner.

Heterogeneity for both inverse-variance weighted (IVW) and MR-Egger methods was evaluated through Cochran’s Q statistics. In light of observed potential heterogeneity, this investigation adopted the random-effects IVW model as the primary analytical approach to explore the causal association between osteoarthritis and hypertension [29]. And other MR methods, such as MR Egger, weighted median, simple mode, and weighted mode, were also employed as reference points. Furthermore, to bolster the assessment of the causal relationship and alleviate the potential influence of horizontal pleiotropy on the outcomes, MR-Egger methods were employed [30, 31]. Moreover, if MR-Egger methods detect horizontal pleiotropy in the genetic instruments, additional steps will be taken to address this issue. These include conducting an outlier test using MR-PRESSO, and subsequently, any outlying SNPs identified in the MR-PRESSO outlier test will be excluded [32]. This process aims to further mitigate the impact of pleiotropic SNPs. All MR analyses were performed using the package “TwosampleMR” [33](version 0.5.6) in R.