Use of oral anticoagulant drugs is associated with carotid intraplaque hemorrhage in atherosclerosis patients: a meta-analysis

We searched multiple electronic databases including Pubmed, Embase, Ovid MEDLINE and Cochrane library for manuscripts that mentioned the relationship between the use of antithrombotic drugs and the higher risk of IPH with language restriction in English, from January 1, 1989 to January 1st, 2019. The search strategy for Ovid database as an example is presented here:

The comprehensive literature search was performed independently by two reviewers to ensure that they satisfied inclusion criteria, any disagreements were resolved by consensus with consultation of experts. We also reviewed and hand searched the reference lists of final included articles to find information pertaining to the topic Search results were limited to human studies, published in English.

Studies included met the following criteria: (1) evaluate the association between antithrombotic drugs use and the risk of carotid IPH; (2) provide the odds ratio (OR) with corresponding confidence interval (CI) or sufficient data to calculate crude OR. Two reviewers independently screened titles and abstracts identified by the preliminary searches to select potentially eligible studies. Studies failing to satisfy both criteria were excluded, and in case of ambiguity as to whether each criterion was fulfilled, abstract under consideration was noted for full-text review. In addition, we excluded articles that were not published in English. In case the results by the same author based on the same population appeared in more than one publication, only the most recent or the most complete one was included in the analysis. Data from review articles, case reports, abstracts, and letters were not included. Consultation with a third reviewer (D. Y. Geng) could be utilized If consensus could not be reached.

Two investigators extracted the following information from each eligible study independently: the name of first author, year of publication, sample type, sample size, population basic characteristics (mean age, sex ratio, BMI), types of drugs, duration of drugs use, numbers of occurrence of carotid IPH, numbers of patients combined history of other disease (hypertension, diabetes mellitus, hyperlipidemia, transient ischemic attacks (TIA) or stroke, ischemic heart disease), the diagnostic approach of IPH, the ORs with corresponding 95% CI and adjustment for covariates. Differences in data extraction were resolved by consensus, referring back to the original article. The missing or unstated information in the original text would be obtained as much as possible via communicating with the corresponding author. The methodological quality of the included cross-sectional studies was assessed based on the an 11-item checklist recommended by Agency for Healthcare Research and Quality (AHRQ) [20]. An item would score ‘1’ if the answer was ‘YES’, while ‘0’ the answer was ‘NO’ or ‘UNCLEAR’. Article quality was assessed as follows: low quality = 0–3; moderate quality = 4–7; high quality = 8–11. Scores were averaged from the independent assessments of two authors.

All statistical analyses were performed through R software version 3.5.0 (R Foundation for Statistical Computing, Vienna, Austria). OR with 95% CI was used to measure the association between the use of antiplatelet drugs and risk of carotid IPH and the association between the use of anticoagulants and risk of carotid IPH.

The heterogeneity of the estimators of OR was tested by Cochran’s Q test at the P < 0.10 level of significance. We also calculated the quantity I² that describes the percentage variation across studies that is attributed to heterogeneity. An I² value less than 25% was considered low-level heterogeneity, 25% to 50% as moderate-level, and greater than 50% as high-level. When significant heterogeneity was found (I² > 50%), the random-effects model was used for meta-analysis. Otherwise, the fixed-effects model was adopted. And if the clinical heterogeneity was too large, a single research analysis should be used instead. Potential publication bias was evaluated by the Egger’s regression asymmetry test and funnel plot analysis. Sensitivity analysis was performed on the results of association between antiplatelet drugs use with IPH, because Derkson [18] did not provide the adjusted OR of multivariate analysis in their study, we had to calculate the crude OR instead. The sensitivity analysis would be performed to check whether this paper had significantly influenced the outcome. For the analysis of IPH occurrence associated with anticoagulants use, we eliminated two of the four studies. Considering the patient recruited who used anticoagulants also took antiplatelet drugs as dual therapy during treatment in Sun’s study [21]. In the study of Liem [17], they did not have the data of the patient using anticoagulants. Consequently, we only summarized the data of the other two studies [18, 22] for analyzing the association between anticoagulants use and carotid IPH. Limited to the small number of studies included in this part, we did not attempt to do publication bias analysis or sensitivity analysis.

As shown in the flow chart of Fig. 1, our initial literature search identified 308 relevant articles. A number of 145 articles were excluded for duplication, 136 articles were excluded by title and abstract review, including 132 irrelevant topics or animal studies and four reviews, letters, editorials or case reports. Twenty-seven articles were reviewed in full-text for more detailed evaluation. However, 22 articles did not meet the criteria, neither evaluating the association between antithrombotic drugs use and the risk of carotid IPH, nor providing the odds ratio (OR) with corresponding confidence interval (CI). Moreover, there was 1 research based on the ultrasound to diagnose IPH, we ruled it out because we thought the diagnosis method was not accurate enough. As a result, four cross-sectional studies met the predetermined criteria and were included in the meta-analysis. All these four articles provided data on antiplatelet drugs exposure and risk of IPH, and only 2 of them provided data on anticoagulants exposure and risk of IPH.

Table 1 lists all the included studies and their characteristics. All the studies included were published between 2015 and 2018. Four studies, involving a total of 2714 participants with carotid atherosclerotic plaques provided data on the occurrence of IPH after using antiplatelet drugs alone or anticoagulants alone. The number of participants using antiplatelet drugs was 37, 738, 520, 64; and the number of participants using anticoagulants was 116 and 118. In these four studies, the mean age of participants varied from 67 to 72.9 years. Their mean BMI varied from 26.5 to 28.0 kg/m². The types of antiplatelet drugs include aspirin, clopidogrel and dipyridamole, and the types of anticoagulants include coumarin derivate and vitamin K antagonists (VKA). Mean duration of medication and follow-up period varied among the included studies. The four studies included patients who had history of using statin, hypertension, diabetes mellitus, hyperlipidemia, TIA or stroke, and ischemic heart disease. Some of them were regular smokers. Partial information about the exact number of patients with different symptoms and past histories were unknown. In 3 of the studies, the diagnosis of IPH were performed by magnetic resonance imaging (MRI) examinations while the other study used histopathological approaches. Both the 3.0T high-resolution MRI and the 1.5T MRI can detect the IPH accurately based on the high signal in T1-weighted imaging [23]. We tried to contact authors of some papers for the missing or unstated data but did not obtain any extra valid information.

The methodological quality of included cross-sectional studies was assessed using an 11-item checklist recommended by AHRQ. The assessment for each study is provided in Table 2, and their quality scores were 6, 6, 8 and 7 respectively. In summary, 1 study was rated as high quality and the rest 3 were rated as moderate quality.

Four studies, including 2714 patients with carotid atherosclerosis provided data on the risk of occurrence IPH after exposing to antiplatelet drugs. Figure 2a illustrated the forest plot of ORs estimates with 95% CIs from individual studies and overall meta-analysis of all four studies. Using the random effects model, the overall pooled ORs demonstrated no significant association between antiplatelet drugs use and the risk of IPH (OR 1.34; 95% CI 0.68–2.61), with high-level heterogeneity among studies (P = 0.03; I² = 65%, τ² = 0.2830). The tests for funnel plot asymmetry identified no publication bias. Three of the studies provided adjusted ORs, however, while the study by Derkson [18] did not provide the adjusted OR of multivariate analysis. We tried to contact the corresponding author of this paper to obtain the original data. Due to various reasons, the crude OR had to be calculated instead. The crude OR was 0.50, and the 95% CI was 0.21–1.18. Through the sensitivity analysis (Fig. 2b), omitting this study, the pooled OR of the other 3 studies still suggested no significant association between antiplatelet drugs exposure and the risk of IPH (OR 1.75; 95% CI 0.92–3.33). Thus, the crude OR did not change the outcome statistically.

Considering the patient used anticoagulants also took antiplatelet drugs as dual therapy in Sun’s study [21], and the absence of data analysis in Liem’s study [17] about the patient using anticoagulants, we ruled out these two studies before performing the meta-analysis. Two studies, including 2534 patients with carotid atherosclerosis provided data on the risk of developing IPH after exposing to anticoagulants. Figure 3 illustrated the forest plot of ORs estimates with 95% CIs from individual studies and overall meta-analysis of the two studies. Using the fixed effect model, the pooled ORs demonstrated a significant association between anticoagulants and the occurrence of IPH (OR 1.95; 95% CI 1.16–3.30), with low-level heterogeneity among studies (P = 0.92; I² = 0, τ² = 0). We did not attempt to do an analysis of publication bias or subgroup analyses as there were only 2 studies included in the analysis.