Risk of major adverse cardiovascular events among traditional Chinese medicine users: a cohort study with a subgroup analysis of male patients with sexual dysfunction
Riesgo de eventos cardiovasculares adversos mayores entre los usuarios de medicina tradicional china: un estudio de cohortes con un análisis de subgrupos en pacientes varones con disfunción sexual

2.1 Data source

This retrospective cohort study utilized data from Taiwan’s National Health Insurance Research Database (NHIRD), which has provided de-identified, comprehensive healthcare data for nearly the entire Taiwanese population since 1995. The NHIRD encompasses both TCM and Western medicine, allowing for detailed analyses of integrative healthcare utilization. In this study, TCM specifically refers to CHM and does not include other TCM modalities such as acupuncture or manual therapy. In 2016, the database transitioned from the ICD-9-CM to the ICD-10-CM diagnosis codes. This study was approved by the Ethics Committee of China Medical University Hospital, Taiwan (approval number: CMUH109-REC2-031). Due to privacy regulations, all personal data within the NHIRD is encrypted, thus obviating the need for informed consent.

2.2 Study design

Between 2001 and 2018, 2 million Taiwanese individuals were randomly selected from the NHIRD and categorized into TCM users and non-TCM users. TCM users were defined as patients who received TCM for more than 30 days, whereas individuals treated for less than 30 days were classified as non-TCM users. The index date was defined as the first recorded instance of TCM use for the TCM group, while it was randomly assigned for the non-TCM group.

To ensure comparability between groups, a stepwise selection process was applied. Patients who did not meet the inclusion criteria were excluded before conducting propensity score matching (PSM). After exclusion, 1:1 PSM was performed based on age, comorbidities (such as hypertension, diabetes, hyperlipidemia, coronary artery disease, and chronic kidney disease), medication use (including PD5EIs, SSRIs and nitrates) and index year. This matching process resulted in a balanced cohort of male TCM users (n = 144,114) and male non-TCM users (n = 144,114), effectively minimizing potential confounding factors.

From this matched cohort, male patients diagnosed with SD were identified for a subgroup analysis to assess whether the cardiovascular risks observed in SD patients were specific to their condition or reflected a general trend among male TCM users (Fig. 1).

Fig. 1.Flowchart illustrating the recruitment of patients with MSD receiving TCM or non-TCM treatment from the NHIRD in Taiwan from 2001–2018. MACEs: major adverse cardiovascular events; MSD: male sexual dysfunction; NHIRD: National Health Insurance Research Database; SD: sexual dysfunction; TCM: traditional Chinese medicine.

2.3 Primary outcomes and confounders

The primary outcome was the incidence of MACEs, including ischemic stroke, hemorrhagic stroke, acute coronary syndrome, heart failure, and malignant arrhythmia after the index date [23]. Primary outcomes and baseline comorbidities were identified via disease codes, requiring documentation from at least two outpatient visits or one hospitalization. Potential confounders included comorbidities such as hypertension, diabetes, coronary artery disease, atherosclerosis, prostate diseases, alcohol-related illnesses, obesity, depression, anxiety, hyperlipidemia, chronic kidney disease, illicit drug dependence or abuse and nicotine dependence [24]. Comparisons of medication use included PDE5Is, amide drugs, SSRIs and nitrates. The detailed disease and Anatomical Therapeutic Chemical (ATC) codes are provided in Supplementary Table 1.

2.4 Statistical analysis

Propensity scores were estimated using logistic regression, using TCM as the dependent variable and baseline characteristics as independent variables, to create balanced cohorts of TCM and non-TCM users. Covariate balance was assessed post-matching. Baseline demographics and comorbidities were compared using Chi-square tests, with continuous variables presented as means ± standard deviations and categorical variables presented as counts and percentages. Differences between groups were evaluated via unpaired t-tests and standardized mean differences (SMD), with SMD values <0.1 indicating no significant difference [25]. Hazard ratio (HR) and 95% confidence interval (CI) were calculated via Cox proportional hazards regression, adjusting for age, comorbidities and medication use. Sensitivity analyses were performed to examine the consistency of results across patient subgroups stratified by age, comorbidities and SD status, aiming to determine whether the observed associations remained robust under different conditions. The Kaplan-Meier method was used to estimate the cumulative incidence of MACEs, with survival curves generated using R software. All other statistical analyses were conducted using Statistical Analysis System (SAS) software, version 9.4 (SAS Institute Inc., Cary, NC, USA), with statistical significance set at p < 0.05.

3.1 Comparison of baseline characteristics between non-TCM and TCM cohorts, including the SD subgroup

The baseline demographics and comorbidities of the matched cohorts (144,114 TCM users and 144,114 non-TCM users) are summarized in Table 1. The mean age was similar between the groups: 41.40 ± 12.82 years for non-TCM users and 41.08 ± 12.81 years for TCM users. The follow-up periods were also comparable, with 9.27 ± 5.21 years for non-TCM users and 9.71 ± 5.22 years for TCM users. Within this matched cohort, 4959 participants were diagnosed with SD, comprising 1921 non-TCM users and 3038 TCM users. Across both the overall cohort and the SD subgroup, no significant differences were observed between the groups in terms of comorbidities or medication use between the groups (all SMD values <0.1).

3.2 HR and 95% CI for MACEs in the overall cohort, including patients with and without SD

Table 2 presents the HR and incidence rates of MACEs in the overall cohort, which includes both patients with and without SD. After adjusting for age, comorbidities and medication use, the incidence rates of MACEs were 9.93 and 9.08 per 1000 person-years in the TCM and non-TCM groups, respectively. Compared with non-TCM users, the TCM users presented a higher risk of MACEs (adjusted hazard ratio (aHR) 1.07, 95% CI (1.04, 1.09), p < 0.001), along with a 1.24-fold in risk of ischemic stroke (aHR 1.24, 95% CI (1.20, 1.28), p < 0.001). Conversely, TCM users, compared to non-TCM users, had significantly lower risks of hemorrhagic stroke (aHR 0.80, 95% CI (0.76, 0.85), p < 0.001), along with lower risks for acute coronary syndrome (aHR 0.93, 95% CI (0.88, 0.98), p = 0.007) and malignant arrhythmia (aHR 0.91, 95% CI (0.83, 0.99), p = 0.031). No significant difference was observed for heart failure (aHR 0.95, 95% CI (0.91, 1.00), p = 0.048). Fig. 2 illustrates the cumulative incidence of MACEs and specific cardiovascular events in the overall cohort, comparing TCM and non-TCM users. The Kaplan-Meier analysis showed a significantly higher cumulative incidence of MACEs (Fig. 2A) and ischemic stroke (Fig. 2B) in TCM users (log-rank test, p < 0.001), but a lowerer incidence (log-rank test, p < 0.001) for hemorrhagic stroke (Fig. 2C). No significant differences were found for acute coronary syndrome (Fig. 2D), heart failure (Fig. 2E), or malignant arrhythmia (Fig. 2F).

Fig. 2.Cumulative incidence curves of cardiovascular outcomes among TCM and non-TCM users. (A) major adverse cardiovascular events (MACEs). (B) ischemic stroke. (C) hemorrhagic stroke. (D) acute coronary syndrome. (E) heart failure. (F) malignant arrhythmia. The analysis includes all patients, encompassing both those with and without SD. SD: sexual dysfunction; TCM: traditional Chinese medicine.

3.3 Stratification analysis of MACEs associated with TCM use

Supplementary Table 2 represents the stratified analysis, and demonstrates that TCM users had a significantly greater risk of MACEs compared to non-TCM users among the individual aged 50–59 years (aHR 1.09, 95% CI (1.04, 1.13), p < 0.001) and 60–69 years (aHR 1.09, 95% CI (1.04, 1.14), p < 0.001). Similarly, an increased risk was noted among patients with hypertension (aHR 1.14, 95% CI (1.07, 1.22), p < 0.001) and hyperlipidemia (aHR 1.23, 95% CI (1.12, 1.35), p < 0.001). Conversely, patients with chronic kidney disease exhibited a significantly reduced risk of MACEs (aHR 0.84, 95% CI (0.72, 0.98), p = 0.026). Additionally, a slightly elevated risk of MACEs was observed in TCM users concurrently taking SSRIs (aHR 1.24, 95% CI (1.02, 1.50), p = 0.028). No significant associations were found with PDE5Is, amide drugs, illicit drug dependence or nicotine dependence.

3.4 HR and incidence rates for MACEs among male patients with different types of SD (subgroup analysis)

Table 3 presents the HR and incidence rates for MACEs among male patients with different types of SD. In the subgroup of 4959 male patients with SD, TCM users had a significantly higher risk of MACEs than non-TCM users (aHR 1.23, 95% CI (1.05, 1.43), p = 0.009). Further analysis by SD type showed that this increased risk was primarily observed in patients with organic SD (aHR 1.24, 95% CI (1.06, 1.45), p = 0.006). In contrast, no significant difference was found among those with non-organic SD (aHR 0.58, 95% CI (0.19, 1.82), p = 0.353).

4.1 Association between TCM use and MACEs: main findings and subgroup analysis

This study is the first large-scale retrospective cohort study to examine the association between TCM use and MACEs. In the overall cohort, the aHR for MACEs was 1.07 among TCM users compared with non-TCM users, even after adjusting for age, comorbidities and medication use. While this association remained statistically significant, the relatively small effect size suggests that minor residual differences in baseline characteristics or unmeasured confounders may have influenced the result.

Selection bias may have contributed to this association. Some TCM users may have sought treatment for conditions unrelated to cardiovascular or SD, reflecting differences in baseline health status, undiagnosed comorbidities and healthcare-seeking behaviors. These differences could independently influence cardiovascular risk, indicating that the observed association may partly reflect baseline health disparities rather than a direct causal effect of TCM use. Although our statistical adjustments aimed to minimize these biases, residual confounding cannot be entirely excluded.

However, selection bias alone may not fully explain the observed association. Our subgroup analysis revealed that the increased risk of MACEs was particularly pronounced among patients with SD who used TCM (aHR 1.23), especially in those with organic SD (aHR 1.24). Given that SD itself is an early marker of vascular dysfunction and cardiovascular disease [3], these findings suggest that the increased cardiovascular risk may not solely result from baseline health differences, but could also involve potential biological effects associated with TCM.

Although our study cannot establish causality, it highlights the need for further investigation into the interplay between TCM use, SD severity and cardiovascular health. To minimize confounding, we employed PSM and multivariate Cox regression models to account for major cardiovascular risk factors, such as hypertension and hyperlipidemia. Despite these adjustments, the risk of MACEs remained higher in the TCM group compared with the non-TCM group, suggesting that additional factors, including potential pharmacological effects of CHMs, may contribute to this association.

Further research incorporating clinical biomarkers, pharmacodynamic assessments, and prospective study designs is needed to clarify whether the observed association represents a clinically meaningful difference or is partially attributable to inherent limitations in observational data.

4.2 Potential biological explanations for the observed association

Several hypotheses may explain this association. One possibility is that certain kidney-yang tonifying herbs commonly used for SD treatment may influence cardiovascular function. Herbs, such as Epimedium Leaf (Yin-Yang-Huo, Epimedii folium), Morinda Root (Ba-Ji-Tian, Morindae officinalis radix), Desert-living Cistanche (Rou-Cong-Rong, Cistanchis herba) and Songaria Cynomorium Herb (Suo-Yang, Cynpomorii herba), are frequently prescribed in TCM to enhance kidney-yang and sexual function [26, 27, 28, 29, 30, 31]. These herbs are traditionally described as having warming and energizing properties, which may enhance metabolic activity and stimulate adrenergic responses. Some studies suggest that they may also modulate vascular function, potentially affecting blood pressure regulation and sympathetic nervous system activity [19, 32, 33, 34].

Additionally, some kidney-yang tonics may exert androgenic effects, resembling testosterone, which is known to influence cardiovascular function [30, 35, 36]. These androgenic effects could lead to increased blood pressure and vasoconstriction [37, 38], aligning with our findings that patients with hypertension or hyperlipidemia using TCM faced a greater risk of MACEs. Certain CHMs, such as Ginseng Root (Ren-Shen, Ginseng radix et rhizome), have been reported to cause a temporary increase in blood pressure and palpitations when used as aphrodisiacs [33, 36, 39]. These short-term hemodynamic changes suggest a possible mechanism through which kidney-yang tonics could influence cardiovascular outcomes, particularly in individuals with pre-existing cardiovascular conditions. However, the long-term cardiovascular impact of these herbs remains uncertain, highlighting the need for further research to elucidate their potential effects on vascular health.

4.3 Other cardiovascular considerations associated with TCM use

Beyond the increased risk of MACEs, other cardiovascular considerations related to TCM use warrant further discussion. One particularly notable finding was the potential interaction between SSRIs and kidney-tonifying Chinese herbs concerning cardiovascular outcomes. Our analysis showed that TCM users concurrently taking SSRIs exhibited a slightly elevated risk of MACEs (aHR 1.24, 95% CI (1.02, 1.50), p = 0.028). Although this association reached statistical significance, the lower boundary of the 95% CI was close to 1, suggesting the possibility of a limited clinical effect. SSRIs have previously been reported to be associated with alterations in vascular function and changes in coagulation pathways; however, their potential contribution to cardiovascular risk in combination with TCM use remains uncertain [40, 41, 42]. Further studies are therefore necessary to confirm this relationship and clarify potential biological mechanisms.

Additionally, several observed associations had CI values near 1, warranting cautious interpretation. TCM users with chronic kidney disease showed a modestly reduced risk of MACEs (aHR 0.84, 95% CI (0.72, 0.98), p = 0.026), while slight reductions were noted for acute coronary syndrome (aHR 0.93, 95% CI (0.88, 0.98), p = 0.007) and malignant arrhythmia (aHR 0.91, 95% CI (0.83, 0.99), p = 0.031). These findings may reflect differences in healthcare-seeking behaviors, patient characteristics, or unmeasured confounding rather than the true protective effects of TCM. To clarify their clinical significance, additional research incorporating clinical biomarkers and detailed medication use patterns is required.

4.4 Strengths of the study

This study has several notable strengths, particularly in its methodology and data source. By utilizing Taiwan’s NHIRD, a large and nationally representative database, our study achieves high external validity. The application of PSM and multivariate Cox regression models effectively minimized selection bias by balancing baseline cardiovascular risk factors between TCM users and non-TCM users. Additionally, performing stratified analyses across key subgroups, such as different age groups and patients with or without pre-existing cardiovascular conditions, provided a more nuanced understanding of how TCM use may differentially impact cardiovascular risk.

Furthermore, restricting the study population to males aged 20–70 years ensured greater relevance to the demographic most susceptible to SD, while also reducing potential confounding from extreme age groups [43]. Although this selection approach improved internal comparability, future research should explore whether these findings extend to other demographic groups. Moreover, by excluding patients with mixed organic and non-organic SD, we improved the precision of our analysis in differentiating cardiovascular risks associated with organic versus non-organic SD. This refinement enhances the interpretability of our findings and reduces potential misclassification bias. Collectively, these methodological strengths reinforce the reliability of our findings and enhance the robustness of our conclusions regarding the association between TCM use and cardiovascular risk in male patients with SD.

4.5 Limitations of the study

Despite these strengths, this study has several notable limitations.

First, although we applied PSM and multivariate Cox regression to balance baseline cardiovascular risk factors, residual confounding from unmeasured variables cannot be fully excluded. Taiwan’s NHIRD does not include certain lifestyle factors, such as illegal substance use, nicotine dependence, or out-of-pocket purchases of medications like PDE5Is and amide sprays, which may influence cardiovascular outcomes.

Second, selection bias remains a potential concern. Some TCM users may have had different healthcare-seeking behaviors or underlying health conditions that were not fully captured in our matching process. In particular, some patients may have turned to TCM due to dissatisfaction with conventional treatments, potentially indicating a higher baseline cardiovascular risk among TCM users. Although this scenario could have contributed to the observed association, it is unlikely to fully explain the increased risk of MACEs, given that our subgroup analysis showed a stronger association in patients with SD, particularly organic SD. We adjusted for major cardiovascular risk factors and performed stratification analyses across different age groups (20–29, 30–39, 40–49, 50–59 and 60–69 years). However, differences in comorbidity burden within each age stratum could still exist and partly influence the observed association between TCM use and MACEs incidence. Although we attempted to mitigate this concern by adjusting for individual cardiovascular risk factors, residual confounding from unmeasured comorbidities cannot be entirely ruled out.

Third, the classification of patients into the TCM group was based on an initial exposure criterion of at least 30 days of TCM treatment, rather than the cumulative exposure over the entire study period. This threshold was selected to ensure a meaningful level of engagement with TCM, rather than incidental or sporadic use, and aligns with prior research and pharmacoepidemiological studies in Taiwan, where a 30-day exposure threshold has been commonly applied to evaluate TCM-related risks [44, 45]. Applying an initial treatment period to define exposure is a well-established approach in observational studies, particularly within causal inference frameworks, as it provides a structured method for estimating associations with health outcomes [46]. However, this classification does not distinguish between patients who continued TCM use beyond 30 days and those who discontinued after meeting the minimum criterion. Since our study assessed the association between initial TCM exposure and MACEs in patients with SD, we defined TCM users based on their early treatment status rather than cumulative exposure over time.

The Kaplan-Meier curves illustrate the long-term risk trajectory of patients who initially met the TCM exposure criteria, regardless of ongoing treatment adherence. While this method enables the assessment of long-term cardiovascular risk in individuals with early TCM exposure, its interpretation should consider the limitations of exposure classification in distinguishing short-term from long-term users.

Fourth, although this study specifically defined TCM as CHM, variations in herbal compositions and dosages among individual TCM users do likely introduce heterogeneity. Our study primarily focused on kidney-yang tonifying CHMs commonly used for SD treatment. However, the variability in specific herbal formulations could lead to differences in cardiovascular effects. Although some kidney-yang tonifying CHMs have been suggested to influence cardiovascular function, clinical evidence remains inconclusive. These uncertainties highlight the need for further pharmacological and biomarker-based studies to clarify the cardiovascular safety profile of specific TCM treatments, particularly in high-risk populations.

Fifth, the NHIRD lacks detailed laboratory and clinical data, including testosterone levels, International Index of Erectile Function-5 (IIEF-5) scores and biomarker-based cardiovascular risk assessments [47]. This limitation prevents us from directly assessing SD severity and its cardiovascular implications. However, the NHIRD has been validated for accurately identifying cardiovascular outcomes, reducing the likelihood of misclassification bias [48, 49].

Sixth, the use of TCM could be influenced by underlying health conditions beyond cardiovascular or SD. Although our study focused on TCM use in patients with SD, it is possible that individuals sought TCM treatment for other medical conditions that were not captured in our database. This potential selection bias may have influenced the observed association between TCM use and the risk of MACEs. Future studies should explore the specific clinical indications for TCM prescriptions to better understand their role in cardiovascular outcomes.

Finally, as an observational study, establishing causal relationships between TCM use and cardiovascular risk remains challenging due to the possibility of unmeasured confounding variables. Future studies should incorporate prospective designs, clinical biomarkers, and pharmacokinetic assessments to distinguish between the confounding effects of pre-existing conditions and the potential pharmacological interactions of CHMs.