Increasing risk of hemorrhage following concurrent use of Salvia miltiorrhiza and oral anticoagulants: a retrospective self-controlled case series study

Abstract

Background:

There has been a growing trend of combining traditional Chinese medicine and Western medicine, but the safety of co-prescribing Salvia miltiorrhiza (Danshen) and anticoagulants remains uncertain.

Objectives:

This study aimed to evaluate the bleeding risk following the concurrent prescribing of S. miltiorrhiza and anticoagulants in a real-world setting.

Design:

A self-controlled case series study was conducted.

Methods:

This study used the Chang Gung Research Database to identify adult patients co-prescribed oral anticoagulants and S. miltiorrhiza, and having records of bleeding events. Bleeding events included any bleeding event (gastrointestinal, intracranial, or urogenital bleeding) and major bleeding events (hemorrhages requiring hospitalization or transfusion). Exposure periods were defined as days of concurrent prescribing of S. miltiorrhiza and anticoagulants, while control periods included days of anticoagulant use without S. miltiorrhiza. Conditional Poisson regression was applied to estimate bleeding risk during exposure periods relative to control periods, and results were presented as adjusted incidence rate ratio (aIRR) and 95% confidence interval (95%CI).

Results:

Among 525 patients receiving both medications, 146 experienced bleeding events. The risk of any bleeding increased significantly during days 1–14 (aIRR: 3.98; 95% CI: 3.29–4.77) and days 15–28 (aIRR: 3.99; 95% CI: 3.23–4.79) after concurrent prescribing of S. miltiorrhiza and anticoagulants. Elevated risks of gastrointestinal bleeding (aIRR: 3.79; 95% CI: 2.95–4.53) and intracranial hemorrhage (aIRR: 3.59; 95% CI: 2.79–5.03) were observed within the first 14 days.

Conclusion:

Concurrent use of S. miltiorrhiza and anticoagulants significantly increased bleeding risk, particularly during the first 28 days of coadministration. These findings highlight the need for careful monitoring during the initial period of combined therapy.

Plain language summary

Concurrent use of Danshen (a Chinese herbal medicine) and anticoagulants may increase bleeding risk

Danshen, also known as Salvia miltiorrhiza, is a traditional Chinese medicine that is commonly used for heart and circulation problems. In Taiwan, Danshen is frequently prescribed together with modern Western medications. One group of these medicines, called anticoagulants, are commonly used to prevent blood clots in people with conditions such as atrial fibrillation or heart disease. While anticoagulants are effective, they can increase the risk of bleeding. It has been uncertain whether using Danshen at the same time might make this risk even greater.

In this study, we used a large medical records database to investigate the real-world safety of taking Danshen together with anticoagulants. We included 525 adults who were prescribed both treatments and 146 of them had records of bleeding events. To minimize differences between individuals, we applied a self-controlled case series (SCCS) study design, where each patient served as their own control.

We found that the risk of bleeding was significantly higher during the first 28 days after patients started taking Danshen with an anticoagulant. The risk was especially high for gastrointestinal and intracranial bleeding. On average, the first bleeding event happened about 16 days after starting combined treatment. Importantly, most patients in our study already had a high baseline risk of bleeding, but even after adjusting for other factors, the increased risk remained clear.

These findings suggest that patients who take Danshen alongside anticoagulants should be monitored carefully, particularly during the first few weeks of combined use. This study highlights the need for careful communication about herbal and conventional medicine use.

Keywords

anticoagulants bleeding risk drug interactions self-controlled case series study traditional Chinese medicine

Introduction

More than 40% of the world’s population embraces herbal medicines as part of their healthcare options,¹ and there has been a significant rise in the use of traditional Chinese medicine (TCM) in Taiwan, with nearly 30% of its population having received TCM prescriptions.² Especially, Taiwan’s healthcare landscape encompasses both TCM and Western medicine, with a reported prevalence of approximately 14% in the general population engaging in the concurrent use of both modalities.^3,4

TCM is often perceived by the public as inherently safe and gentle.⁵ However, reports of adverse reactions related to TCM have been increasing steadily, and some studies have highlighted clinically relevant herb–drug interactions.^6,7 This is particularly concerning for patients who may unintentionally take multiple medications, such as the elderly. However, definitive evidence on these interactions remains scarce, and most studies are limited to case reports or animal-based research.⁸

Danshen, the root and rhizome of the perennial herb Salvia miltiorrhiza, is associated with the heart and liver meridians in TCM, and is frequently used in the treatment of cardiovascular diseases, menstrual irregularities, and insomnia.⁹ Among the most commonly prescribed single herbs in Taiwan, S. miltiorrhiza accounts for 16.50% of prescriptions for ischemic stroke.¹⁰ This widespread use increases the likelihood of S. miltiorrhiza being coadministered with Western medications.

For instance, oral anticoagulants are widely prescribed for conditions such as atrial fibrillation and other cardiovascular diseases to prevent thromboembolic events, and they might carry a risk of bleeding.¹¹ S. miltiorrhiza extract has been found to increase the maximum concentration and elimination half-life of warfarin while reducing the clearance and distribution volume of both R- and S-warfarin.¹² In addition, tanshinones, active compounds in S. miltiorrhiza, can inhibit the cytochrome P450 (CYP) enzymes CYP1A1, CYP2C6, and CYP2C11, which mediate the metabolic pathway of warfarin.¹³ When used in combination with warfarin, S. miltiorrhiza has been reported to increase warfarin concentrations by at least 23%.¹³ Based on pharmacokinetic profiles of S. miltiorrhiza’s active constituents, those compounds exhibit half-lives ranging from several hours to nearly 15 h, and hence suggest that the risk of bleeding may be most pronounced in the initial period following coadministration. Furthermore, potential interactions extend to direct oral anticoagulants (DOACs). Although DOACs differ from warfarin in the pharmacodynamic mechanism, they share susceptibility to pharmacokinetic interference. DOACs, such as rivaroxaban and apixaban, are substrates of P-glycoprotein (P-gp) and the cytochrome CYP3A4 enzyme,¹⁴ and studies have demonstrated that the bioactive components of S. miltiorrhiza can inhibit the catalytic activity of various CYP and P-gp.^13,15 These inhibitory effects could increase the systemic exposure of DOACs, thereby potentiating their anticoagulant effects.

The combination of these effects with the anticoagulant activity of either warfarin or DOACs may synergistically impair hemostasis. Despite these theoretical concerns, existing toxicological evidence indicates that S. miltiorrhiza alone may not cause significant bleeding in the absence of anticoagulant therapy.¹⁶ Furthermore, a previous population-based cohort study found that the concurrent use of Chinese herbal medicine and anticoagulants was associated with a lower risk of major bleeding.¹⁷ However, only major bleeding events in the emergency room or hospitalization were included in this study, potentially underestimating clinically relevant bleeding episodes and underscoring a critical gap in our understanding of herb–drug interactions. There remains a lack of large-scale human studies or real-world clinical data confirming the actual risk of bleeding. Therefore, this study aimed to explore the risk of bleeding after the initiation of concurrent prescribing of S. miltiorrhiza and anticoagulants in a real-world setting.

Methods

Data sources

The study was conducted using data from the Chang Gung Research Database (CGRD) between January 2012 and January 2023. The CGRD is a comprehensive repository of patient records from the Chang Gung Memorial Hospital system in Taiwan, encompassing demographic information, International Classification of Diseases, 10th Revision (ICD-10) diagnostic codes, prescription data, and hospitalization records.¹⁸ Importantly, the CGRD includes records of both Western medicine and TCM prescriptions, enabling comprehensive identification of concurrent TCM and anticoagulant use. This study is reported in accordance with the checklist of STROBE statement¹⁹ (see Supplemental Material 1, Table S1).

Study design

This retrospective study used a self-controlled case series (SCCS) design to investigate the risk of bleeding after the initiation of co-prescription of S. miltiorrhiza and anticoagulants. The SCCS design allowed for comparing the incidence of hemorrhage during periods of co-prescription with that during periods without co-prescription in the same individuals, who acted as their own controls to adjust for time-independent confounding factors, such as genetic variation.^20–22

Study population

The study included patients who met the following criteria: (1) aged ⩾20 years, (2) newly prescribed an anticoagulant (warfarin, dabigatran, rivaroxaban, apixaban, and edoxaban) during the study period, (3) concurrently prescribed S. miltiorrhiza (either as a single herb or in combination formulas), and (4) had a documented bleeding event.

Patients were excluded if they (1) were younger than 20 years, (2) were prescribed oral anticoagulant therapy for less than 7 days, (3) had concurrent use of S. miltiorrhiza and an oral anticoagulant for ⩽1 day, (4) had any record of heparin prescribing before the index date, (5) had incomplete demographic information, or (6) had received any anticoagulant prescription in the 6 months before their first anticoagulant prescription during the study period. A 6‑month washout window prior to the index date (defined as the date of the first anticoagulant prescription during the study period) to ensure that only new anticoagulant users were included in the cohort. Accordingly, S. miltiorrhiza was considered concurrently prescribed if the exposure periods of S. miltiorrhiza and an oral anticoagulant overlapped for more than 1 day.

Documented bleeding events included any bleeding event and major bleeding events. Any bleeding event was defined as the first documented diagnosis of gastrointestinal, intracranial, or urogenital bleeding, identified through at least two outpatient department visits or at least one emergency room visit. Major bleeding was defined as hemorrhage requiring at least one hospital admission or red blood cell transfusion, as identified from inpatient department and emergency room records. All bleeding events were identified using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) and ICD-10-CM diagnostic codes (detailed codes are listed in Supplemental Material 2, Table S1).

Definition of exposure and control periods

The co-prescription period was defined as the duration during which both S. miltiorrhiza and anticoagulants were concurrently prescribed. The start date of the exposure period was defined as the first date on which the prescription durations of both medications overlapped, regardless of the sequence of initiation (i.e., whether S. miltiorrhiza was added to existing anticoagulants or vice versa). The exposure period ended on the last day of the overlapping prescription duration. This definition ensures that the analysis strictly captures the risk associated with the drug–drug interaction window. The exposure period was categorized into four risk periods: days 1–14 after co-prescription, days 15–28 after co-prescription, days 29–56 after co-prescription, and the post-exposure period, defined as a 28-day interval following the 56th day of co-prescription to monitor any residual effects after discontinuation (Figure 1). The follow-up time interval other than the exposure period was defined as the control period.

Figure 1.

Pictorial representation of the self-controlled case series study design.

Confounding factors

The time-dependent confounders in our analysis were age, concurrent medications (antiplatelets, steroids, proton pump inhibitors, and nonsteroidal anti-inflammatory drugs (NSAIDs)), and comorbidities such as hypertension and heart failure (a complete list of ICD codes is available in Supplemental Material 2, Table S1). These variables were selected a priori based on their known or suspected influence on the risk of bleeding and their potential to vary over the study period.²³ Besides, to assess baseline bleeding risk, we calculated the HAS-BLED score for each patient using clinical and laboratory data from the CGRD. The score includes the following components: hypertension, abnormal renal and/or liver function, history of stroke, history of bleeding or predisposition to bleeding, labile international normalized ratio (INR), age ⩾65 years, and concomitant use of drugs (aspirin or NSAIDs) or alcohol.²³ Each component was assigned one point according to the original HAS-BLED definition. Hypertension, stroke, renal/liver dysfunction, and bleeding history were identified using ICD diagnosis codes. Concomitant drug use was determined from prescriptions within 30 days prior to the index date, and alcohol use was identified from structured clinical records. Labile INR was defined according to the NICE guideline as having either: (1) two INR values >5, or one INR value >8 within the past 6 months; or (2) two INR values <1.5 within the past 6 months.²⁴

Statistical analysis

Conditional Poisson regression was used to estimate the adjusted incidence rate ratios (aIRRs) and 95% confidence interval (95% CI) of bleeding events during the exposure period relative to the control period. For each individual, the observation time was partitioned into exposure and control intervals. The natural logarithm of the number of days in each interval [log(days)] was included as an offset term to standardize for differences in follow‑up duration across intervals. In addition, to account for the possibility that bleeding events within the same individual may be correlated across different time intervals, we employed robust variance estimation with individual-level clustering. Age was treated as a time-varying covariate to capture the temporal increase in baseline bleeding risk within each individual. Comorbidities were coded as time-varying covariates: individuals were considered unexposed to a given comorbidity prior to its first recorded diagnosis; from diagnosis onward, the comorbidity indicator remained present for the remainder of follow-up. Concomitant medications (antiplatelets, systemic steroids, proton-pump inhibitors [PPIs], and NSAIDs) were adjusted for as time-varying covariates, as these medications are prescribed or adjusted in response to acute changes in health status and care intensity. In addition, absolute event rates were calculated as the number of bleeding events per 1000 person-years (PY) for each risk interval to provide a measure of absolute risk. To evaluate the robustness of our findings across different patient characteristics, we performed subgroup analyses stratified by age (<75 and ⩾75 years), sex, and type of anticoagulant (warfarin vs DOACs).

The following sensitivity analyses were conducted to ensure the robustness of the findings: (1) Events that occurred in the pre-exposure period, defined as a 14-day interval immediately preceding the start of co-prescription, were excluded from the primary analysis to ensure that the observed outcomes were attributable to drug exposure rather than the underlying condition prompting the prescription.²⁵ (2) Only the first bleeding event for each patient was included in the analysis to mitigate potential bias from recurrent events.

Results

Characteristics of the study cohort and bleeding events

A total of 79,426 patients receiving oral anticoagulants were identified from the CGRD between January 1, 2012 and January 31, 2023. Finally, 146 patients who were co-prescribed S. miltiorrhiza and anticoagulants and had records of bleeding events during the observation period were identified (Figure 2).

Figure 2.

Flow diagram of patient selection and exclusions.

The study population demonstrated a balanced gender distribution, with a mean age in the early 70s (Table 1). DOACs were more commonly prescribed than warfarin, with dabigatran (n = 69, 47.26%) being the most frequently used anticoagulant. Cardiovascular conditions, particularly hypertension (n = 125, 85.62%) and atrial fibrillation (n = 120, 82.19%), were the predominant comorbidities. Most patients were receiving multiple concurrent medications, with antiplatelets (n = 115, 78.77%) being the most common concomitant therapy. Notably, more than 90% of patients had a high baseline risk of bleeding, as indicated by HAS-BLED scores of 2 or greater.

Table 1.

Baseline characteristics of patients with bleeding events who were co-prescribed anticoagulants and Salvia miltiorrhiza.

Characteristics	Patients (n = 146, 100%)
Sex
Male	76 (52.05)
Female	70 (47.95)
Age on the index date, median (IQR), years	73.56 (13.53)
Follow-up duration, median (IQR), days	675.0 (430.0)
Anticoagulants
Warfarin	42 (28.77)
Dabigatran	69 (47.26)
Rivaroxaban	31 (21.23)
Apixaban	33 (22.60)
Edoxaban	23 (17.75)
Comorbidities
Hypertension	125 (85.62)
Rheumatoid arthritis	5 (3.42)
Heart failure	59 (40.41)
Renal disease	63 (43.15)
Peripheral arterial occlusive disease	17 (11.64)
Myocardial infarction	5 (3.42)
Atrial fibrillation	120 (82.19)
Dementia	19 (13.01)
Peptic ulcer disease	89 (60.96)
Mild liver disease	62 (42.47)
Paraplegia and hemiplegia	19 (13.01)
Hemophilia/coagulopathy	11 (7.53)
Concomitant medications
Antiplatelets	115 (78.77)
Steroids	96 (65.75)
Proton pump inhibitors	98 (67.12)
Nonsteroidal anti-inflammatory drugs	79 (54.11)
HAS-BLED risk score
0	1 (0.68)
1	10 (6.85)
⩾2	135 (92.47)

HAS-BLED risk score, hypertension, abnormal renal/liver function, stroke, bleeding history or predisposition, labile INR, elderly, drugs/alcohol concomitantly score; IQR, interquartile range.

The study population predominantly consisted of elderly patients, with a mean age of 73.18 ± 9.26 years, and the majority were in their seventies at the time of their first bleeding event. The duration of anticoagulant therapy was generally longer than the period of concurrent use of S. miltiorrhiza and anticoagulants. Bleeding events typically occurred within the first few weeks after the initiation of combination therapy (Table 2). A detailed comparison between patients with and without bleeding events is provided in Supplemental Material 2, Table S2. The majority of patients (85.6%) experienced a single bleeding event, with recurrent events occurring in only 14.4% of cases (see Supplemental Material 2, Table S3 for details).

Table 2.

Characteristics of bleeding events in patients co-prescribed anticoagulants and Salvia miltiorrhiza.

Characteristics	Patients ( n = 146, 100%)
Age at first bleeding event, median (IQR), years	73.11 (13.43)
Duration of drug use, median (IQR), days
Anticoagulants	309.5 (495.0)
Co-prescribing of anticoagulants and S. miltiorrhiza	56.0 (142.0)
Duration from co-prescribed anticoagulants and S. miltiorrhiza to the bleeding event, median (IQR), days	16 (19)
Duration from anticoagulant prescription to event, mean (SD), days
Warfarin	1308.38 (1367.33)
Dabigatran	729.55 (801.70)
Rivaroxaban	579.29 (581.51)
Apixaban	491.58 (631.97)
Edoxaban	871.83 (936.61)
Duration from concomitant medication prescription to event, mean (SD), days
Antiplatelets	2485.53 (1916.13)
Steroids	2749.26 (2079.91)
Proton pump inhibitors	1945.92 (1842.25)
Nonsteroidal anti-inflammatory drugs	3312.53 (2163.06)

IQR, interquartile range; SD, standard deviation.

Association between any bleeding event and concurrent S. miltiorrhiza and anticoagulant use

The absolute incidence rate of any bleeding was 174.17 per 1000 PY during the non-exposure control periods. This rate increased markedly to 218.99 per 1000 PY during the first 14 days of co-prescription (Table 3), corresponding to an absolute risk difference (ARD) of 44.82 per 1000 PY, or approximately 4.5 additional bleeding events per 100 PY. The aIRR for any bleeding events were 3.98 (95% CI: 3.29–4.77), 3.99 (95% CI: 3.23–4.79), and 3.88 (95% CI: 3.23–4.45) at days 1–14 after co-prescription, days 15–28 after co-prescription, and the entire exposure period, respectively (Figure 3).

Table 3.

Adjusted incidence rate ratios of bleeding events during risk periods after coadministration of S. miltiorrhiza and anticoagulants.

Period	Events	Followed PY	Absolute event rates (per 1000 PY)	Crude incidence rate ratio	Adjusted incidence rate ratio* (95% CI)
Intracranial hemorrhage
Control period	8	492.98	16.23	1.00	1.00
Exposure period
1–14 days	14	84.94	164.82	10.16	3.59 (2.79–5.03)
15–28 days	12	70.35	170.57	10.51	3.65 (2.84–4.94)
1–56 days	32	216.44	147.85	9.11	2.86 (1.01–3.72)
Post-exposure (28 days)	11	169.88	64.75	3.99	2.89 (1.04–3.64)
Gastrointestinal bleeding
Control period	39	587.23	66.41	1.00	1.00
Exposure period
1–14 days	31	153.93	201.39	3.03	3.79 (2.95–4.53)
15–28 days	22	149.60	147.06	2.21	3.60 (2.82–4.28)
1–56 days	71	434.29	163.49	2.46	3.43 (2.67–3.99)
Post-exposure (28 days)	21	307.87	68.21	1.03	3.82 (3.07–4.65)
Any bleeding
Control period	51	292.82	174.17	1.00	1.0
Exposure period
1–14 days	44	200.93	218.99	1.26	3.98 (3.29–4.77)
15–28 days	35	185.91	188.26	1.08	3.99 (3.23–4.79)
1–56 days	103	549.30	187.51	1.08	3.88 (3.23–4.45)
Post-exposure (28 days)	24	401.85	59.72	0.34	2.81 (2.37–3.39)

Adjusted by age and drug use (steroids, PPIs, antiplatelets, and NSAIDs).

NSAIDs, nonsteroidal anti-inflammatory drugs; PY, person-year.

Figure 3.

Forest plot of aIRRs for bleeding events.

The risk of intracranial hemorrhage remained high at days 1–14 after co-prescription (aIRR: 3.59; 95% CI: 2.79–5.03; ARD: 148.59 per 1000 PY) and days 15–28 after co-prescription (aIRR: 3.65; 95% CI: 2.84–4.94; ARD: 154.34 per 1000 PY) compared to the control periods. For gastrointestinal bleeding, the aIRR was 3.79 (95% CI: 2.95–4.53; ARD: 134.98 per 1000 PY) at days 1–14 after co-prescription, and 3.60 (95% CI: 2.82–4.28; ARD: 80.65 per 1000 PY) at days 15–28 after co-prescription. No major bleeding events were observed during any episode of the exposure period.

Subgroup analyses

The results of the subgroup analyses are presented in Supplemental Material 2, Table S4. The association between concurrent use of S. miltiorrhiza and anticoagulants and increased bleeding risk remained consistent across key subgroup analyses, including age, gender, and anticoagulant types.

Sensitivity analyses

When the 14-day pre-exposure period was excluded to minimize potential bias from baseline bleeding risks, the results remained robust, demonstrating consistently elevated risks of bleeding during different episodes of the exposure period (Supplemental Material 2, Table S5). The results showed some deviations from the primary analysis when considering only the first bleeding event per patient. An elevated risk of gastrointestinal bleeding was only observed during the first 14 days after co-prescription (aIRR: 2.02; 95% CI: 1.14–2.91), and there was no significant risk increase during the exposure episodes for any bleeding event (Supplemental Material 2, Table S6).

Discussion

This study demonstrated a significantly increased risk of bleeding events, particularly within the first 28 days of combination therapy with S. miltiorrhiza and oral anticoagulants in a real-world setting. The observed median time to first bleeding event of 16 days further supports a potential association between S. miltiorrhiza initiation and bleeding events. Notably, this elevated risk was consistently observed across different bleeding types and time frames, underscoring the importance of close clinical monitoring during this period. These findings are particularly significant as 92.47% of the study population had a HAS-BLED score of ⩾2, reflecting an already high baseline risk of bleeding complications.

The timing of the elevated bleeding risk observed in our study may be attributed to the pharmacokinetics of S. miltiorrhiza’s active compounds. Lipophilic components such as cryptotanshinone and tanshinone IIA exhibit relatively longer half-lives (up to 14.8 h), whereas hydrophilic components like danshensu and protocatechuic acid are eliminated more rapidly.²⁶ This pharmacokinetic profile supports the classification of early risk windows in our analysis and helps explain the temporal association between co-prescription and bleeding events.

Beyond pharmacokinetics, the increased risk of bleeding may also be attributed to S. miltiorrhiza’s pharmacodynamic effects and its interactions with anticoagulants. S. miltiorrhiza contains multiple bioactive compounds, including tanshinones and phenolic acids (e.g., danshensu and salvianolic acids), which have demonstrated effects on blood circulation and coagulation.^27,28 The mechanisms behind these interactions are complex and multifaceted, with pharmacokinetic studies suggesting diverse pathways through which S. miltiorrhiza can significantly influence the metabolism and efficacy of various anticoagulants. For instance, S. miltiorrhiza extract has been shown to significantly influence the pharmacokinetics of warfarin by increasing its maximum concentration and elimination half-life while reducing the clearance and distribution volume of both R- and S-warfarin isomers.¹² Regarding the interaction with DOACs, the pharmacokinetic mechanisms appear complex and theoretically conflicting. Clinical evidence indicates that S. miltiorrhiza extract can induce intestinal CYP3A4 activity in healthy volunteers, which could theoretically accelerate the metabolism and reduce the plasma concentrations of CYP3A4.²⁹ By contrast, another study suggests that lipophilic tanshinones can inhibit P-gp and CYP enzymes, which could conversely increase systemic exposure to DOACs.¹⁵ Our subgroup analysis showed the risk of any bleeding was significantly elevated not only in warfarin users (aIRR 5.25) but also in DOAC users (aIRR 3.85) in the first 14 days. Although DOACs are generally considered to have a lower baseline risk of bleeding compared to warfarin, our findings indicated similar results across types of anticoagulants. Furthermore, S. miltiorrhiza possesses intrinsic antiplatelet and antithrombotic properties.^28,30 The additive effect of this antiplatelet activity on a background of systemic anticoagulation provides an explanation for the increased bleeding risk observed across both anticoagulant classes.

Beyond pharmacokinetics, S. miltiorrhiza also exerts direct cardiovascular effects. Previous studies have demonstrated that it dilates coronary arteries, increases coronary blood flow, and inhibits platelet adhesion and aggregation.^28,30 These effects, while potentially beneficial in certain clinical contexts, may further increase the risk of bleeding when S. miltiorrhiza is used concurrently with anticoagulants. Despite its potential risks, S. miltiorrhiza offers significant therapeutic benefits in various cardiovascular conditions. Recent studies have shown that S. miltiorrhiza, and its active component tanshinone IIA can significantly reduce endothelial inflammation through inhibiting cyclooxygenase-2 and modulating the TNF-α/NF-κB signaling pathway.^31,32 Tanshinone IIA has also demonstrated cardioprotective effects by inhibiting ferroptosis and apoptosis.³³ A recent network meta-analysis evaluating the clinical effects of S. miltiorrhiza-containing injections in patients with acute myocardial infarction found that the injection of sodium tanshinone IIA sulfonate was effective in reducing mortality and enhancing cardiac function. However, the authors emphasized that careful monitoring for bleeding events was essential during treatment.³⁴

Our findings contrast with previous observational studies on the bleeding risks associated with the concurrent use of TCM herbs and anticoagulant drugs. A retrospective cohort study suggested that the use of Chinese herbal medicine, including S. miltiorrhiza (Danshen), Panax notoginseng, Glycyrrhiza uralensis, and Curcuma longa might decrease the risk of major bleeding events in patients taking anticoagulants (hazard ratio: 0.87; 95% CI: 0.805–0.940). However, those Chinese herbal medicines differ from each other in pharmacological properties and interaction mechanisms, and there was a neutral effect (adjusted hazard ratio: 0.948; 95% CI: 0.851–1.056) on major bleeding events when only S. miltiorrhiza was considered due to the limited sample size.¹⁷

Our findings have important clinical implications for the management of patients taking oral anticoagulants. While S. miltiorrhiza has demonstrated cardiovascular benefits, clinicians should be mindful of its potential to increase the risk of bleeding in the first 2 weeks after being prescribed alongside anticoagulation therapy.^26,30,35 This caution is particularly relevant given S. miltiorrhiza’s widespread use for treating cardiovascular conditions,^27,28 which increases the likelihood of its concurrent use with anticoagulants.

This concern may extend beyond S. miltiorrhiza, as similar bleeding risks have been observed with other herbal medicines. For example, Chan et al.³⁶ found that Ginkgo biloba combined with anticoagulants increased the risk of hemorrhage, particularly in elderly patients. Similarly, Chan et al.³⁷ reported that herbal consumption affected the time in the therapeutic range of warfarin therapy in patients with atrial fibrillation. In addition, studies have shown that both prescription and over-the-counter (OTC) medications, including herbal medicines, can significantly impact anticoagulation control.^36,37 In Taiwan, S. miltiorrhiza-containing products are widely available over the counter and are commonly used without a prescription, making it difficult for physicians to fully ascertain a patient’s complete medication profile. These observations highlight the need for vigilant assessment of herb–drug interactions in patients taking anticoagulants.

Our study has several strengths. The use of an SCCS design, which has been successfully used in previous pharmacoepidemiologic studies on drug safety, allowed for the elimination of time-invariant confounders. To our knowledge, this is the first study to apply the SCCS design to evaluate the bleeding risk associated with concurrent use of S. miltiorrhiza and oral anticoagulants. Given that our outcome of interest was acute bleeding events, the SCCS design offers methodological advantages over cohort studies by better accounting for within-person variability and minimizing confounding. Our study also provided more comprehensive exposure assessments and broader outcome definitions than previous research, allowing for a better capture of the real-world clinical impact of herb-drug interactions. Moreover, this study adopted CGRD, one of the largest and most comprehensive multi-institutional electronic medical record databases in Taiwan, thereby enhancing the generalizability of our findings to the broader population.

Limitations

Several limitations should be acknowledged. First, as with most observational studies using health care databases, we could not verify actual medication adherence, and our exposure assessment was based on dispensing records rather than direct observation. Second, the CGRD only captures prescriptions dispensed within the Chang Gung hospital system; however, S. miltiorrhiza-containing products are also available OTC and may be purchased without a prescription. As a result, some true periods of concurrent use (e.g., OTC S. miltiorrhiza taken during ongoing anticoagulant therapy) could have been misclassified as unexposed control time. This non-differential misclassification of exposure would bias our aIRRs toward the null, suggesting that the true association between concurrent S. miltiorrhiza and anticoagulant use and bleeding may be stronger than our estimates. Third, our definition of “any bleeding” was restricted to gastrointestinal, intracranial, and urogenital sites. While other bleeding manifestations (such as hemoptysis or epistaxis) are clinically relevant, they were not included in our primary outcome to minimize measurement error and focus on the most impactful adverse events. Previous research indicates that gastrointestinal and intracranial hemorrhage are the most frequent anticoagulant-related bleeding events requiring hospitalization.³⁸ Furthermore, validation studies have demonstrated that diagnosis codes for these major sites possess significantly higher validity. By contrast, codes for other bleeding sites are often prone to underreporting and demonstrate lower precision.³⁹ By focusing on these reliable and predominant anatomical sites, we aimed to ensure the diagnostic specificity and clinical representativeness of our findings, although this definition may conservatively underestimate the total burden of minor bleeding events. Fourth, we were unable to account for unmeasurable confounders, such as changes in diet, alcohol consumption, and use of other herbal medicines. Regarding confounding by indication and reverse causality, we acknowledge that our analysis did not explicitly adjust for specific prescribing indications. To mitigate concerns regarding reverse causality (i.e., S. miltiorrhiza being prescribed in response to clinical conditions related to bleeding risk), we conducted a sensitivity analysis excluding the 14-day pre-exposure period. This standard pharmacoepidemiologic approach effectively isolates the effect of concurrent drug exposure from confounding factors that may have prompted the prescription. The robust and consistent results of this pre-exposure exclusion analysis with the primary analysis provide reassurance that confounding by indication is unlikely to be a major driver of our findings. Fifth, as the majority of our study population had HAS-BLED scores ⩾2, limiting the generalizability of our findings to patients with lower baseline bleeding risk, this reflects the clinical reality that patients prescribed anticoagulants often possess multiple risk factors for bleeding, such as advanced age and hypertension. However, this implies that our findings may not be fully generalizable to populations with a low baseline bleeding risk (e.g., younger patients without comorbidities). Consequently, the observed high relative risk should be interpreted within this high-risk cohort. Sixth, the SCCS method relies on the assumption that the occurrence of an event does not influence subsequent exposure. In clinical practice, a bleeding event may lead to treatment discontinuation. However, our data indicated that recurrent bleeding events were uncommon, occurring in only 14.4% of patients. The restriction to the first bleeding event in our sensitivity analysis inevitably reduced statistical power, leading to wider CIs. Despite this limitation, the risk signal for gastrointestinal bleeding remained significant, indicating that the association is unlikely to be driven solely by event-dependent changes in exposure.

Further research into the specific mechanisms of interaction between S. miltiorrhiza compounds and different anticoagulants could help develop more targeted monitoring strategies and risk assessment tools. Future studies should also assess whether the increased bleeding risk applies to patients with lower baseline risk (e.g., HAS-BLED <2), who were underrepresented in our study.

Conclusion

In conclusion, this SCCS study demonstrated that the concurrent use of S. miltiorrhiza and oral anticoagulants was associated with a significantly increased risk of bleeding events, particularly within the first 28 days of coadministration. The risk was most pronounced for gastrointestinal and intracranial bleeding, suggesting that healthcare providers should carefully assess the risk–benefit ratio and consider enhanced monitoring when administering S. miltiorrhiza to patients receiving anticoagulation therapy. Future research should focus on developing optimal monitoring strategies and risk assessment tools for S. miltiorrhiza and oral anticoagulant coadministration, thereby providing important evidence to enhance patient safety in the context of integrated TCM and Western medicine.

Supplemental Material

sj-pdf-1-taw-10.1177_20420986261446497 – Supplemental material for Increasing risk of hemorrhage following concurrent use of Salvia miltiorrhiza and oral anticoagulants: a retrospective self-controlled case series study

Supplemental material, sj-pdf-1-taw-10.1177_20420986261446497 for Increasing risk of hemorrhage following concurrent use of Salvia miltiorrhiza and oral anticoagulants: a retrospective self-controlled case series study by Jian-An Liao, Teng-Chou Chen, Pony Yee Chee Chai, Tien-Hsing Chen and Yung-Chih Chen in Therapeutic Advances in Drug Safety

Supplemental Material

sj-pdf-2-taw-10.1177_20420986261446497 – Supplemental material for Increasing risk of hemorrhage following concurrent use of Salvia miltiorrhiza and oral anticoagulants: a retrospective self-controlled case series study

Supplemental material, sj-pdf-2-taw-10.1177_20420986261446497 for Increasing risk of hemorrhage following concurrent use of Salvia miltiorrhiza and oral anticoagulants: a retrospective self-controlled case series study by Jian-An Liao, Teng-Chou Chen, Pony Yee Chee Chai, Tien-Hsing Chen and Yung-Chih Chen in Therapeutic Advances in Drug Safety

Footnotes

Acknowledgements

We would like to extend our gratitude to research assistant Xue-Ming Lin for his valuable technical assistance with the analyses.

Declarations

ORCID iDs

Jian-An Liao

Teng-Chou Chen

Yung-Chih Chen

Supplemental material

Supplemental material for this article is available online.

References

WHO global report on traditional and complementary medicine 2019. Geneva: World Health Organization, 2019.

Chen

Kung

, et al. Use frequency of traditional Chinese medicine in Taiwan. BMC Health Serv Res 2007; 7: 26.

Chen

Lai

Chen

, et al. Concurrent Use of Conventional Drugs with Chinese Herbal Products in Taiwan: a population-based Study. J Tradit Complement Med 2013; 3: 256–262.

Ulbricht

Chao

Costa

, et al. Clinical evidence of herb-drug interactions: a systematic review by the natural standard research collaboration. Curr Drug Metab 2008; 9: 1063–1120.

Liu

Chuang

Lam

, et al. Safety surveillance of traditional Chinese medicine: current and future. Drug Saf 2015; 38: 117–128.

Yang

, et al. Herb-drug interactions: a literature review. Drugs 2005; 65: 1239–1282.

Chen

Iqbal

, et al. Interactions between traditional Chinese medicine and western drugs in Taiwan: a population-based study. Comput Methods Programs Biomed 2015; 122: 462–470.

Chan

TY.

Interaction between warfarin and danshen (Salvia miltiorrhiza). Ann Pharmacother 2001; 35: 501–504.

Zhou

Zuo

Chow

MS.

Danshen: an overview of its chemistry, pharmacology, pharmacokinetics, and clinical use. J Clin Pharmacol 2005; 45: 1345–1359.

10.

Hung

Wang

, et al. Chinese herbal products for ischemic stroke. Am J Chin Med 2015; 43: 1365–1379.

11.

Di Minno

Frigerio

Spadarella

, et al. Old and new oral anticoagulants: Food, herbal medicines and drug interactions. Blood Rev 2017; 31: 193–203.

12.

Chan

Yeung

, et al. The effects of Danshen (Salvia miltiorrhiza) on warfarin pharmacodynamics and pharmacokinetics of warfarin enantiomers in rats. J Pharm Pharmacol 1995; 47: 402–406.

13.

Yeung

JH.

Inhibition of warfarin hydroxylation by major tanshinones of Danshen (Salvia miltiorrhiza) in the rat in vitro and in vivo. Phytomedicine 2010; 17: 219–226.

14.

Lee

, et al. The increased risk of bleeding due to drug-drug interactions in patients administered direct oral anticoagulants. Thromb Res 2020; 195: 243–249.

15.

Chen

Tian

DD.

Salvia miltiorrhiza roots against cardiovascular disease: consideration of herb-drug interactions. Biomed Res Int 2017; 2017: 9868694.

16.

Mei

. Handbook of pharmacology and clinical application of modern Chinese medicine, 3rd ed. Beijing: China Press of Chinese Medicine, 2016, p. 994.

17.

Hsu

Lin

Kao

, et al. Concurrent use of Chinese herbal medicine and anticoagulants may reduce major bleeding events. PLoS One 2022; 17: e0271965.

18.

Shao

Chan

Kao Yang

, et al. The Chang Gung Research Database: a multi-institutional electronic medical records database for real-world epidemiological studies in Taiwan. Pharmacoepidemiol Drug Saf 2019; 28: 593–600.

19.

von Elm

Altman

Egger

, et al. [The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies]. Rev Esp Salud Publica 2008; 82: 251–259.

20.

Petersen

Douglas

Whitaker

Self-controlled case series methods: an alternative to standard epidemiological study designs. BMJ 2016; 354: i4515.

21.

Ramsay

Roughead

Ewald

, et al. A self-controlled case series to assess the effectiveness of beta blockers for heart failure in reducing hospitalisations in the elderly. BMC Med Res Methodol 2011; 11: 106.

22.

Whitaker

Farrington

Spiessens

, et al. Tutorial in biostatistics: the self-controlled case series method. Stat Med 2006; 25: 1768–1797.

23.

Pisters

Lane

Nieuwlaat

, et al. A novel user-friendly score (HAS-BLED) to assess 1-year risk of major bleeding in patients with atrial fibrillation: the Euro Heart Survey. Chest 2010; 138: 1093–1100.

24.

Perry

Kemmis Betty

Downes

, et al. Atrial fibrillation: diagnosis and management—summary of NICE guidance. BMJ 2021; 373: n1150.

25.

Requena

Douglas

Huerta

, et al. Impact of pre-exposure time bias in self-controlled case series when the event conditions the exposure: hip/femur fracture and use of benzodiazepines as a case study. Pharmacoepidemiol Drug Saf 2020; 29: 388–395.

26.

Yue

, et al. Pharmacodynamics and pharmacokinetics of Danshen in isoproterenol-induced acute myocardial ischemic injury combined with Honghua. J Ethnopharmacol 2020; 247: 112284.

27.

Chen

Guo

Bao

, et al. The anticancer properties of Salvia miltiorrhiza Bunge (Danshen): a systematic review. Med Res Rev 2014; 34: 768–794.

28.

Cheng

TO.

Cardiovascular effects of Danshen. Int J Cardiol 2007; 121: 9–22.

29.

Qiu

Wang

Zhang

, et al. Effect of danshen extract on the activity of CYP3A4 in healthy volunteers. Br J Clin Pharmacol 2010; 69: 656–662.

30.

Tan

Zhu

YZ.

Salvia miltiorrhiza and ischemic diseases. Acta Pharmacol Sin 2000; 21: 1089–1094.

31.

Zhang

Gao

, et al. Salvia miltiorrhiza and Tanshinone IIA reduce endothelial inflammation and atherosclerotic plaque formation through inhibiting COX-2. Biomed Pharmacother 2023; 167: 115501.

32.

Quan

Wang

Zhang

, et al. The intervention mechanism of Tanshinone IIA in alleviating neuronal injury induced by HMGB1 or TNF-α-mediated microglial activation. Toxicol In Vitro 2024; 101: 105950.

33.

Zou

, et al. Tanshinone IIA confers protection against myocardial ischemia/reperfusion injury by inhibiting ferroptosis and apoptosis via VDAC1. Int J Mol Med 2023; 52: 109.

34.

Chen

Shang

Liu

, et al. Comparative effectiveness and safety of four traditional Chinese medicine injections with invigorating blood circulation, equivalent effect of anticoagulation or antiplatelet in acute myocardial infarction: a Bayesian network meta-analysis. Front Pharmacol 2024; 15: 1400990.

35.

Zhang

Xie

Liao

, et al. A Chinese patent medicine Salvia miltiorrhiza depside salts for infusion combined with conventional treatment for patients with angina pectoris: a systematic review and meta-analysis of randomized controlled trials. Phytomedicine 2017; 25: 100–117.

36.

Chan

Leung

, et al. Risk of hemorrhage associated with co-prescriptions for Ginkgo biloba and antiplatelet or anticoagulant drugs. J Altern Complement Med 2011; 17: 513–517.

37.

Chan

, et al. Effect of herbal consumption on time in therapeutic range of warfarin therapy in patients with atrial fibrillation. J Cardiovasc Pharmacol 2011; 58: 87–90.

38.

Joos

Lawrence

Jones

, et al. Accuracy of ICD-10 codes for identifying hospitalizations for acute anticoagulation therapy-related bleeding events. Thromb Res 2019; 181: 71–76.

39.

Shehab

Ziemba

Campbell

, et al. Assessment of ICD-10-CM code assignment validity for case finding of outpatient anticoagulant-related bleeding among Medicare beneficiaries. Pharmacoepidemiol Drug Saf 2019; 28: 951–964.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.18 MB

0.00 MB

0.31 MB