Association between hypertension and frozen shoulder

Abstract

Observational studies have suggested a potential association between hypertension and frozen shoulder. This study aims to elucidate the causal dynamics of this relationship using a bidirectional 2-sample Mendelian randomization approach. We analyzed data on hypertension (N = 484,598) and frozen shoulder (N = 627,998) from the IEU Open genome-wide association study database. Instrumental variables were carefully selected to ensure robust correlation and minimize linkage disequilibrium. The primary analysis was conducted using the inverse variance-weighted (IVW) method. Additional analyses, including the Weighted Median, MR-Egger, Simple Model, and Weighted Model, were performed to further assess causal effects. We conducted heterogeneity and pleiotropy tests, along with a leave-one-out analysis, to evaluate the stability of our findings. Our analysis revealed that individuals with hypertension have a significantly increased risk of developing frozen shoulder (IVW: odds ratio = 1.749, 95% CI: 1.275–2.399, P = 5.239*10⁻⁴). Conversely, reverse Mendelian randomization analysis showed no significant association between frozen shoulder and the risk of hypertension (IVW: odds ratio = 1.006, 95% CI: 0.991–1.021, P = .426). Our findings confirm that hypertension is associated with an increased risk of frozen shoulder, while the reverse does not hold. Patients with hypertension should be informed of this potential risk and consider preventive strategies against the development of frozen shoulder.

1. Introduction

Frozen shoulder, also known as adhesive capsulitis,^[1] is predominantly characterized by shoulder pain and restricted range of motion.^[2]Epidemiological data indicate a prevalence of 2% to 5%, with a higher incidence observed in women aged 40 to 60 years.^[3] Despite the availability of various treatment options, the optimal therapeutic strategy remains uncertain,^[4] and the underlying etiology of frozen shoulder is not well understood.^[5]

Hypertension, defined by a systolic blood pressure of ≥140 mm Hg, a diastolic blood pressure of ≥90 mm Hg, or control via antihypertensive medications, is a well-established risk factor for numerous medical conditions,^[6] including cardiovascular disease, chronic kidney disease, and stroke, contributing to over 10 million deaths globally each year.^[7] Recent data suggest that 32% of the U.S. population is affected by hypertension, with the prevalence increasing to 45.4% following the American Heart Association’s 2017 redefinition of hypertension criteria.^[8] Although hypertension has been proposed as a potential risk factor for frozen shoulder,^[9] no prior studies have systematically explored the relationship between these 2 conditions. Thus, it is crucial to investigate the genetic causal link between hypertension and frozen shoulder to provide a foundation for clinical treatment and preventive strategies.

Mendelian randomization (MR) is a genetic epidemiological method designed to infer causal relationships from observed associations.^[10] This approach is based on the principle that alleles are randomly assigned and fixed at conception, analogous to the randomization process in controlled trials, which helps to reduce confounding and strengthen causal inference.^[11] In this study, we employed a bidirectional 2-sample MR approach to rigorously examine the potential causal relationship between hypertension and frozen shoulder.

2. Methods

2.1. Study design

Our bidirectional 2-sample MR study followed the framework presented in Figure 1. Genetic variants were employed to examine the causal influence of hypertension on frozen shoulder, as well as to explore reverse causality independently. All MR analyses utilized a 2-sample approach, leveraging publicly available summary statistics, which eliminated the requirement for additional ethical approvals or informed consent.

Figure 1.

Model of the 2-sample Mendelian randomization analysis. MR = Mendelian randomization.

2.2. Data sources

Genetic data for this study were obtained from the IEU Open genome-wide association study (GWAS) database (https://gwas.mrcieu.ac.uk/). Hypertension-related genetic information was sourced from the study by Dönertaş HM et al^[12] which included a cohort of 484,598 individuals. The genetic dataset for frozen shoulder was derived from the research by Green HD et al^[13] encompassing 627,998 cases (Table 1).

Table 1.

Summary of the Genome-wide association study data.

Exposur/outcome	GWAS ID	Number of SNPs	Sample size	Yr
hypertension	ebi-a-GCST90038604	9,587,836	484,598	2021
frozen shoulder	ebi-a-GCST90000513	8423,455	627,998	2021

GWAS = genome-wide association study, SNP = single nucleotide polymorphism.

2.3. Selection of genetic instruments

In this study, a stringent genome-wide significance level of P < 5 × 10⁻⁸ was established. To address the issue of linkage disequilibrium, a clumping algorithm with parameters r² = 0.001 and a 10,000 kb window was employed.^[14] We systematically identified single nucleotide polymorphisms (SNPs) associated with the exposure variable and aligned these with SNPs relevant to the study outcomes. To circumvent confounding effects due to palindromic structures, such SNPs were excluded. The F statistic was meticulously computed for each SNP to ascertain their validity as instruments. SNPs exhibiting weak instrument bias (F < 10) were rigorously excluded, thereby bolstering the integrity and reliability of our Mendelian randomization analyses.

2.4. Statistical analysis

MR analyses were performed using R software (version 4.2.3) with the “TwoSampleMR” package. The inverse-variance weighted (IVW) method was employed as the primary analytical approach to evaluate the causal relationships between exposure factors and outcome variables. To corroborate the robustness of these findings, additional methods including MR-Egger, Weighted Median, Simple Mode, and Weighted Mode were utilized. Unlike IVW, the MR-Egger method accounts for potential pleiotropy by incorporating an intercept term. The Weighted Median approach provides consistent estimates even when up to 50% of the instrumental variables (IV) are invalid.^[15]Although Simple Mode and Weighted Mode have less statistical power compared to IVW, they contribute to the overall assessment of the reliability and stability of the results.^[16]

2.5. Sensitivity analysis

In this study, pleiotropy was assessed using the intercept of the MR-Egger regression, with a P value of <.05 indicating the presence of horizontal pleiotropy. The Cochran Q test was employed to evaluate heterogeneity among the IV, with a P value >.05 suggesting minimal heterogeneity, thus favoring the use of the fixed-effect IVW method. In cases where significant heterogeneity was detected (P value < .05), the random-effects IVW method was utilized.^[17] A thorough sensitivity analysis was conducted using the “leave-one-out” approach, where each SNP was sequentially excluded to assess the impact of the remaining SNPs, ensuring the robustness of the findings. Additionally, funnel plots were used to examine the symmetry of scatter distributions, aiding in the identification of any outlying SNPs.

3. Results

3.1. Causative effect of hypertension on frozen shoulder

3.1.1. Selection of genetic instruments

In this study, we initially evaluated 38,095 single nucleotide polymorphisms (SNPs) associated with hypertension. Linkage disequilibrium analysis resulted in the exclusion of 37,822 SNPs, reducing the number to 273. Further analysis indicated the loss of 42 additional SNPs when correlated with frozen shoulder. During data reconciliation, we excluded 31 palindromic SNPs with intermediate allele frequencies. The remaining IV all had F statistics exceeding 10, demonstrating their robustness. This meticulous selection process culminated in the identification of 200 SNPs for investigating the relationship between hypertension and frozen shoulder (supplementary table1, Supplemental Digital Content 1). A comprehensive examination of the human genotype-phenotype association database confirmed the absence of other SNPs relevant as risk factors in this context.

3.1.2. Causal relationship between hypertension and frozen shoulder

Using the IVW method, we identified a significant causal relationship between hypertension and the incidence of frozen shoulder, with an odds ratio (OR) of 1.749 (95% CI: 1.275–2.399, P = 5.239*10⁻⁴). This finding was consistently supported across various complementary analytical approaches, including MR-Egger, Weighted Median, Simple Mode, and Weighted Mode, all of which aligned in effect direction with the IVW results. Notably, the MR-Egger and Weighted Median methods showed statistical significance with P values below .05 (Table 2). The association between hypertension and an increased risk of frozen shoulder was further corroborated visually through scatter plots (Fig. 2) and forest plots (Fig. 3), reinforcing the robustness of these findings.

Table 2.

MR regression results of the 5 methods.

Methods	β	SE	OR(95%CI)	P
MR Egger	1.083	0.415	2.955 (1.310–6.662)	9.701E–03
Weighted median	0.578	0.209	1.78 (1.183–2.688)	5.772E–03
Inverse variance weighted	0.559	0.161	1.749 (1.275–2.399)	5.239E–04
Simple mode	0.544	0.527	1.724 (0.613–4.847)	.303
Weighted mode	0.544	0.356	1.724 (0.858–3.461)	.127

Figure 2.

Scatter plot of the 2-sample Mendelian randomization analysis.

Figure 3.

Forest plot of the 2-sample Mendelian randomization analysis.

3.1.3. Sensitivity analysis

MR-Egger regression produced an intercept of −4.4577*10⁻³ (P = .172), suggesting minimal horizontal pleiotropy. Although Cochran’s Q test detected some heterogeneity, this did not undermine the results obtained through the IVW method(IVW: Cochran Q = 300.616, P = 5.693 × 10⁻⁶). This confirms the robust causal link between hypertension and frozen shoulder and validates the IV used in this study. The reliability of these results was further supported by a sensitivity analysis using the leave-one-out approach, which demonstrated that excluding any individual SNP did not significantly affect the causal estimate (Fig. 4). Additionally, the symmetrical distribution observed in the funnel plot (Fig. 5) attests to the absence of significant bias in the analysis.

Figure 4.

Sensitivity analysis to investigate the possibility that the causal association was driven by a unique single nucleotide polymorphism in frozen shoulder.

Figure 5.

Funnel plot of the 2- sample Mendelian randomization analysis.

3.2. Causative effect of frozen shoulder on hypertension

3.2.1. Selection of genetic instruments

Initial analysis identified 155 SNPs associated with frozen shoulder.After excluding 150 SNPs due to linkage disequilibrium, 5 SNPs remained. Subsequent analysis confirmed the association of these SNPs with frozen shoulder, with no further exclusions. During data reconciliation, 1 palindromic SNP with intermediate allele frequencies was excluded. The remaining IV demonstrated F statistics significantly exceeding 10, ranging from 32 to 159. Consequently, 4 SNPs were selected as IV for analyzing the relationship between frozen shoulder and hypertension (Supplementary table 2, Supplemental Digital Content 2). A thorough search of the human genotype-phenotype association database revealed no additional SNPs relevant as potential risk factors.

3.2.2. Causal relationship between frozen shoulder and hypertension

Inverse Mendelian randomization analysis revealed no statistically significant association between frozen shoulder and hypertension (IVW: OR = 1.006, 95% CI: 0.991–1.021, P = .426) (Table 3). Although Cochran’s Q test detected some heterogeneity, this did not undermine the results obtained through the IVW method(IVW: Cochran Q = 10.810, P = .013). Additionally, the pleiotropy assessment using the MR-Egger intercept (3.228 × 10⁻³) showed no significant pleiotropic effects (P = .182).

Table 3.

MR regression results of the 5 methods.

Methods	β	SE	OR(95%CI)	P
MR Egger	−1.904E–02	1.343E–02	0.981 (0.956–1.007)	.292
Weighted median	−1.030E–03	4.713E–03	0.999 (0.990–1.008)	.827
Inverse variance weighted	5.870E–03	7.375E–03	1.006 (0.991–1.021)	.426
Simple mode	8.625E–04	7.605E–03	1.001 (0.986–1.016)	.917
Weighted mode	−7.824E–04	4.752E–03	0.999 (0.990–1.009)	.880

4. Discussion

Utilizing GWAS data, this study rigorously explored the causal dynamics between hypertension and frozen shoulder through a bidirectional 2-sample Mendelian randomization approach. The findings suggest a unidirectional causal relationship in which hypertension increases the risk of frozen shoulder, while no reciprocal effect on hypertension incidence was observed from frozen shoulder. This aligns with previous observational studies suggesting hypertension as a potential risk factor for frozen shoulder. Notably, Daniel C. Austin et al^[18]observed a higher prevalence of blood pressure medication usage among patients with frozen shoulder (33.1%, 95% CI: 25.9%–41.2%) compared to the national average (21.6%, 95% CI: 19.8%–23.4%). Furthermore, another study involving 1983 patients reported a higher incidence of systemic arterial hypertension in individuals with frozen shoulder (10.5%) compared to those without (6%).^[19]

However, Kiera Kingston et al.^[20] found a lower likelihood of hypertension in patients with frozen shoulder (OR, 0.93; P = .006) in a study of hospitalized patients from 2004 to 2014, highlighting the potential for confounding factors and reverse causality in retrospective analyses. MR, an innovative epidemiological method that uses genetic variants as IV for exposures, strengthens causal inferences and mitigates the influence of confounders.^[21]

The present 2-sample bidirectional Mendelian randomization study genetically confirmed a unidirectional causal effect of hypertension on an elevated risk of adhesive capsulitis, with no evidence of horizontal pleiotropy or reverse causality. These genetic findings are biologically plausible and can be explained by multiple underlying pathophysiological pathways linking hypertension to capsular fibrosis and shoulder joint dysfunction.

First, hypertension-induced microvascular ischemia contributes to adhesive capsulitis pathogenesis. Chronic hypertension induces systemic microvascular endothelial dysfunction, arteriosclerosis,^[22] and reduced perfusion of shoulder periarticular soft tissues. Persistent local ischemia and hypoxia impair fibroblast repair function, trigger abnormal fibroblast proliferation and phenotypic transformation, and cause excessive extracellular matrix and collagen deposition.^[23] Such microangiopathy-mediated tissue remodeling ultimately leads to shoulder capsular thickening, fibrosis, and adhesion, consistent with previous findings that hypertensive microcirculatory disturbance drives soft tissue fibrotic hyperplasia.^[24]

Second, hypertension-related chronic low-grade systemic inflammation facilitates capsular fibrosis. Hypertension is accompanied by persistent elevation of circulating pro-inflammatory cytokines (TNF-α, IL-6, CRP),^[25] which infiltrate shoulder periarticular tissues and activate local inflammatory responses.^[26] Chronic inflammation disrupts shoulder capsular tissue remodeling balance, promotes fibroblast fibrogenic activity, and induces progressive capsular adhesion.^[27] Mechanically, hypertension-activated angiotensin II further amplifies inflammatory cascades and soft tissue fibrotic deposition.^[28]

Third, sympathetic nervous system overactivation in hypertension aggravates shoulder capsular fibrosis. Sustained sympathetic hyperactivity in hypertensive patients increases circulating catecholamine levels, which induce local microvascular vasoconstriction and worsen tissue ischemia.^[29] Additionally, sympathetic overactivation directly modulates fibroblast biological behaviors, upregulates fibrogenic genes, and accelerates collagen synthesis and capsular fibrosis.^[30] The synergistic crosstalk between sympathetic overactivity and systemic inflammation further exacerbates shoulder tissue pathological damage and increases adhesive capsulitis susceptibility.

Notably, our reverse MR analysis excluded the causal effect of adhesive capsulitis on hypertension, indicating that the above pathophysiological pathways are unidirectional. Hypertensive pathological changes (microvascular injury, systemic inflammation, and sympathetic hyperactivity) are systemic preconditioning factors for shoulder capsular fibrosis, rather than secondary inflammatory or stress responses induced by local joint lesions. Collectively, these mechanistic explanations strongly support our genetic causal inference, making the association between hypertension and adhesive capsulitis biologically credible and clinically interpretable.

Consistent with the majority of prior research, our results indicate that hypertension is a significant risk factor for frozen shoulder. This suggests an increased need for screening and early intervention in hypertensive patients to potentially improve outcomes in cases of frozen shoulder. Notably, several limitations of the present study should be acknowledged. First, all genetic data used in this 2-sample MR study were derived from European ancestry cohorts, which may restrict the generalizability of our conclusions to other ethnic populations. Differences in genetic background and gene-environment interactions may alter the magnitude and direction of genetic causal effects in non-European populations. Given that our research team is based in China and focuses on population-related metabolic and musculoskeletal disorders, we strongly emphasize the necessity of future large-scale East Asian and multi-ethnic GWAS data to replicate and validate the current causal findings. Further trans-ethnic Mendelian randomization analyses will help clarify the ethnic specificity of the hypertension-adhesive capsulitis causal pathway and verify the cross-population applicability of our results, which is essential for promoting the ethnic generalization and clinical translation of relevant precision prevention strategies. Additionally, although the current GWAS sample size is substantial, future studies should aim to expand it. Lastly, the hypertension data did not differentiate between diastolic and systolic blood pressure, indicating the need for more detailed research to validate these findings.

Author contributions

Conceptualization: Danyun Hua.

Project administration: Danyun Hua.

Validation: Danyun Hua.

Data curation: Yingying Chen, Lu Wang, Na Li.

Investigation: Yingying Chen.

Software: Lu Wang, Xiang Xu.

Methodology: Na Li, Xiang Xu.

Resources: Na Li.

Writing – original draft: Danyun Hua, Lu Wang, Na Li.

Writing – review & editing: Danyun Hua.