Bidirectional two-sample Mendelian randomization analysis identifies causal associations between cardiovascular diseases and frozen shoulder

Frozen shoulder (FS) is a fibroproliferative disorder characterized by pain and progressive restriction of shoulder motion [1]. Despite its self-limiting nature, with spontaneous recovery typically occurring within 1–2 years, research suggested that numerous FS-related symptoms could persist in 20–50% of patients, encompassing stiffness and pain [2, 3]. The treatment of FS emphasizes physical therapy, pharmacological relief, and potential surgery. Meanwhile, preventive measures are crucial for alleviating patient symptoms and reducing the socio-medical burden [4]. Certain preventive measures can regulate the occurrence and progression of FS; thus, identifying novel FS risk factors and interventions is imperative.

Cardiovascular diseases (CVDs), primarily encompassing coronary artery disease (CAD), heart failure (HF), myocardial infarction (MI), and stroke, have emerged as the leading causes of morbidity and mortality in the global population [5]. World Health Organization statistics indicated an annual death toll of about 17.8 million due to CVD, contributing to roughly 30% of the total global fatalities [6]. Recent studies have demonstrated a close association between various CVDs and FS. A clinical retrospective study involving 262 FS patients reported that the overall risk of FS was more than three times higher in CAD patients [7]. FS patients have a 1.22 times higher risk of experiencing a stroke compared to the healthy control group [8]. Similarly, a large cohort study indicated that FS patients face an elevated risk of MI compared to the general population [9]. Overall, the existing findings in the observational studies section demonstrate the correlation between CVDs and FS. Despite the association, the mentioned studies have not adequately distinguished the causal relationship between CVDs and FS, nor have they addressed the evidence gap that may result from unmeasured confounding or reverse causation. Therefore, caution is still needed in interpreting these relationships.

Observational studies face limitations in establishing causality due to the inability to control interventions, confounding, and bias. In recent years, in order to overcome the limitations of observational studies, Mendelian randomization (MR) has emerged as a novel approach to explore causal association between etiology and diseases [10, 11]. Based on the genome-wide association study (GWAS), MR uses natural genetic variation to simulate randomized controlled trials (RCTs) and uses genetic variation as instrumental variables (IVs) to assess the causal relationship between exposures and outcomes. In MR studies, single-nucleotide polymorphisms (SNPs) are utilized as IVs to estimate the causal associations between exposures and target outcomes [12]. These SNPs adhere to the random distribution principle of genetic variation, effectively mitigating the potential impacts of confounding factors and reverse causality, given that genetic variations manifest prior to the onset of the disease [13]. Meanwhile, the advancing GWASs have furnished robust and dependable IV support for MR study.

Although Lv et al. have reported the causal relationship between CVD and FS using MR methods, their study was limited to the association between stroke and FS and did not analyze the reverse causal relationship between them [14]. Therefore, the current study employs newly released GWAS data to conduct a two-sample bidirectional MR study, aiming to investigate the presence of a direct causal association between multiple CVDs and FS, providing a theoretical basis for clinical practice. To our knowledge, this is the first two-sample MR study endeavoring to unveil the causality between multiple CVDs and FS. By circumventing certain limitations of observational studies in the process of causal inference, this study holds the potential to offer a clearer perspective for us to delve deeper into understanding this association.

We utilized a bidirectional two-sample MR study for the first time to investigate the causal relationships between CVDs and FS. The results indicated that some types of CVDs were related to the risk of FS. Specifically, genetic susceptibility to stroke and IS was significantly associated with FS risk, and this association remains significant even after FDR correction. Nevertheless, we lacked substantial evidence to establish a link between AF, CAD, HF, MI, and FS. However, we did not find sufficient evidence to support connections between AF, CAD, HF, MI, with FS. These findings exhibited robustness and consistency across various sensitivity analyses, further enhancing the reliability of our study results.

Previous studies have shown that stroke increases the risk of FS. This observation aligns with the conclusion of our study. Relevant studies indicated that the most common complication one year after stroke was shoulder pain, accounting for approximately 40% [30]. Most shoulder pain occurs within 2 months after onset, with 50% of them attributed to FS [31]. Over time, the risk of developing FS in participants 2 to 6 months after stroke occurrence increases four times [32]. Furthermore, MR studies have also verified a causality between IS and FS [14]. Firstly, stroke might trigger an inflammatory response in the shoulder joint, stimulate fibrosis of the muscles around the shoulder area, and subsequently elevate the risk of developing FS. Stroke results in cerebral ischemia or hemorrhage, causing the release of chemical compounds called inflammatory mediators when brain cells are damaged or die. These compounds include cytokines and inflammatory proteins, which then activate immune cells and inflammatory cells, initiating an inflammatory response [33, 34]. A broad consensus exists that the progression of FS is closely linked to inflammation and fibrosis. Studies have indicated a significant elevation of inflammatory cytokines such as interleukin (IL)-1α, IL-1β, tumor necrosis factor (TNF)-α, and cyclooxygenase-2 in the joint capsules of FS patients [35]. Extensive presence of inflammatory cells like T cells, B cells, macrophages, and mast cells has been widely detected in supraspinatus muscle interval samples from FS patients [36]. Secondly, central nervous system lesions following a stroke often lead to limb paralysis, progressively increasing the risk of joint effusion and muscle adhesions, ultimately contributing to the occurrence of FS [37, 38]. Finally, Some medications used in stroke treatment, especially those affecting blood coagulation and circulation, may have adverse effects on joints and muscles, thereby increasing the risk of developing FS. These medications may lead to blood vessel narrowing, affecting local blood supply and causing damage to joint tissues [39]. Additionally, drugs may modulate the immune response in joints through the immune system, leading to inflammation and damage. Side effects of medications may include discomfort or pain in muscles and joints, while a decrease in mobility may restrict normal shoulder joint movement, creating unfavorable conditions for the development of FS [40]. This complex interplay may make patients more prone to developing shoulder periarthritis.

It is noteworthy that there is currently no consensus on whether FS increases the risk of stroke. One study indicated that the risk of stroke in FS patients was 1.46-times (95% CI 1.32–1.62; P < 0.001) that of the control group, and even after adjusting for confounding factors, the risk remained elevated at 1.22-times(95% CI 1.06–1.40; P = 0.002) [8]. Subsequently, another follow-up study based on a large propensity score-matched population found no association between FS and an increased risk of stroke (OR [95% CI] = 0.93 (0.83–1.04), P = 0.178) [41]. This is consistent with our study findings. We mainly attribute this result to the substantial imbalance in demographic characteristics and distribution of vascular risk factors between the FS and non-FS groups in the study mentioned above [8]. Compared to the non-FS group, the FS group had a significantly older age, and the prevalence of diabetes and hyperlipidemia was 1.5 times higher. Despite adjustments in the multivariate regression analysis, the marked imbalances in these potential confounding variables could present challenges in effectively addressing these factors.

Interestingly, this study found no clear causal relationship between MI, CAD, and FS, which differs from previous study results and presents an intriguing trend. One study revealed that the risk of FS in CAD patients was 3.128-times (95% CI, 1.428–7.019; P = 0.005) that of non-FS patients [7]. On the one hand, shoulder pain in CAD patients may result in muscle contraction due to heart disease, thereby restricting shoulder joint movement. On the other hand, insufficient coronary artery blood supply and impaired cardiac function in CAD patients affect local vascular circulation, resulting in tissue ischemia and hypoxia, which can also cause shoulder pain [42]. Another follow-up study on the risk of FS patients indicated that over a 3-year follow-up period, the incidence risk of MI increased by 1.49-times (95% CI, 1.25–1.77; P < 0.001) [9]. They indicated that FS patients may exhibit elevated levels of C-reactive protein (CRP) and alterations in related immune complexes [43]. However, CRP plays a significant role in CVDs and is closely associated with their prognosis [44]. We speculate that the reasons for the inconsistency in study results may include the following two aspects: First, CAD, MI, and FS share some pathogenic factors, such as age, hyperlipidemia, and diabetes. CAD and MI may indirectly affect FS through these common pathogenic factors. MR studies may adopt more stringent methods, potentially better controlling confounding factors, while cross-sectional studies may yield inconsistent results due to the incomplete elimination of confounding factors. Second, MR studies focus more on the causal relationship's time sequence, while cross-sectional studies mainly reflect the correlation at the observed time points, potentially leading to differences in study outcomes. Importantly, the existing literature concerning this area of research is limited. Therefore, a cautious evaluation of these study results remains necessary, and future investigations should be conducted to attain a more comprehensive understanding of the interrelationships between MI, CAD, and FS.

However, our study results currently do not support a causal relationship between AF, HF, and FS, and no related research reports have been found to date. We speculate that AF and HF are CVD involving disruptions in heart structure and function, while FS is an inflammatory disorder around the joints. Since they impact distinct physiological systems and mechanisms, establishing a direct causal relationship is difficult.

We found that stroke increases the risk of developing FS, similar to Lv et al.'s study. However, we focus on the bidirectional causal relationship between six CVDs and FS, going beyond the singular relationship between stroke and FS. The results confirm that stroke increases the risk of FS, but there is no evidence that FS affects stroke. This deepens the understanding of the impact of FS on different CVDs. Although no causal relationship was found between FS and the other four CVDs, observational studies on the relationship between CAD, MI, and FS provided stronger evidence. Overall, by broadening the scope of our study, we ultimately confirmed the causal relationship between stroke and FS, providing a comprehensive and in-depth perspective on understanding the relationship between CVDs and FS.

The roles of various CVDs in the occurrence and progression of FS are complex and potentially interactive. Although our MR analysis was able to account for their interactive effects and assess their causality from a genetic perspective, our study has inherent limitations. Firstly, the results of this study are products of statistical analysis, and some correlations between CVDs and FS have not been reported, lacking theoretical support. Further, additional fundamental and clinical research is necessary to substantiate our observations. Secondly, our data were obtained from public databases that do not permit subgroup analyses for specific factors (such as age, gender, etc.). Future investigations into the correlation between CVDs and FS should account for the overall impact of these factors. Lastly, we exclusively utilized genetic data from individuals of European ancestry, limiting the generalizability of our study findings.

In summary, stroke is a high risk factor for FS. Orthopedic physicians should prioritize shoulder care for stroke patients, promptly assessing potential restrictions in shoulder range of motion to mitigate FS risk. However, further research is needed to validate and extend these findings.