Antidepressant Use and Risk of Venous Thromboembolism: A Systematic Review and Meta-Analysis.

PURPOSE
Studies provided conflicting results on whether antidepressant use increased the risk of venous thromboembolism (VTE). Our aim was to examine the association between antidepressant use and the risk of VTE.


METHODS
Pubmed, Embase, and the Cochrane Library were searched up to March 13, 2018. Case-control studies and cohort studies that examined the association between antidepressant use and the risk of VTE, deep vein thrombosis or pulmonary embolism were included. Several subgroup analyses and sensitivity analyses were conducted. GRADE approach was used to assess the quality of evidence.


RESULTS
Nine studies (six case-control studies and three cohort studies) were included. Overall, antidepressant use may be associated with an increased risk of VTE (OR 1.27, 95% CI 1.09 to 1.49); however, no association was observed in studies with low risk of bias (OR 1.27, 95% CI 0.84 to 1.92). No association between selective serotonin reuptake inhibitor use and VTE risk was detected in the overall analysis (OR 1.10, 95% CI 0.90 to 1.34) and in subgroup analysis of studies with low risk of bias. Tricyclic antidepressant may be associated with an increased VTE risk (OR 1.26, 95% CI 1.02 to 1.57), and the quality of evidence was rated as very low by GRADE approach; however, no association was observed when we only included studies with low risk of bias.


CONCLUSIONS
There was no association between selective serotonin reuptake inhibitor use and VTE risk. Tricyclic antidepressant may be associated with an increased VTE risk, but the quality of evidence was very low.


INTRODUCTION
Venous thromboembolism (VTE) comprises deep vein thrombosis (DVT) and pulmonary embolism (PE) (1). VTE may be fatal in the acute phase, and may lead to many long-term complications (2)(3)(4). Many risk factors for VTE have been identified, such as pregnancy, surgery, cancer and so on (1). VTE may also be associated with some drugs.
As the incidence of psychiatric disorders increases, the use of psychotropic drugs (antipsychotic drugs and antidepressant drugs) has increased dramatically (5)(6)(7). Two meta-analyses showed that antipsychotic drugs may be associated with an increased risk of VTE (without examine the relationship between antidepressant use and VTE risk) (8,9); however, whether a relationship exists between antidepressant exposure and VTE is unclear. Some cases report the occurrence of VTE in patients with antidepressant use have been published in the past decade (10)(11)(12).
Some observational studies have been designed to evaluate the association between antidepressant and VTE; however, their results are controversial. Some of them found an association between antidepressant use and VTE risk (13,14), whereas others not (15,16). In addition, several studies indicated that different classes of antidepressant may be associated with different risk of VTE (14,15).
The objective of this systematic review and metaanalysis was to examine the association between antidepressant use and the risk of VTE, and to ascertain the VTE risk associated with different type of antidepressant.

Search strategy
A systematic search of Pubmed, Embase, and the Cochrane Library was conducted from inception to March 13, 2018, using the following combined Medical Subject Headings (MeSH) and free text words: (antidepressant or antidepressive or "monoamine oxidase inhibitors" or "serotonin reuptake inhibitors" or tricyclic or amitriptyline or clomipramine or dosulepin or dothiepin or doxepin or imipramine or maprotiline or amoxapine or desipramine or nortriptyline or protriptyline or trimipramine or lofepramine or moclobemide or _________________________________________ isocarboxazid or phenelzine or tranylcypromine or iproniazid or citalopram or escitalopram or fluoxetine or paroxetine or fluvoxamine or sertraline or venlafaxine or desvenlafaxine or duloxetine or milnacipran or levomilnacipran or trazodone or mirtazapine or nefazodone or vilazodone or bupropion or mianserin or tianeptine or viloxazine or maprotiline or agomelatine or vortioxetine or reboxetine) and (thromboembolism or "venous thromboembolism" or "venous thrombosis" or "pulmonary embolism" or "pulmonary thromboembolism" or "deep vein thrombosis"). In addition, the reference lists of included studies and review articles were screened for potential eligible studies. Included and excluded studies were collected following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (17).

Study selection
Two reviewers independently performed the study selection, including screening titles and abstracts, and retrieving full texts of studies for details. Disagreements were resolved by discussion or through consultation with other authors. Casecontrol studies and cohort studies that examined the association between antidepressant use and the risk of VTE, DVT or PE were included. Only studies provided odds ratio (OR), risk ratio (RR), or hazard ratio (HR) with 95% confidence intervals (CI), or provided data allowing the calculation of the estimate (OR, RR, or HR) and 95% CI were included. No language restriction was applied.
Data extraction and quality assessment Following information were extracted from the included studies: author name, year of publication, study design, data source, population, number of patients, definition of antidepressant exposure, outcomes, and adjusted/matched factors. Two reviewers independently assessed the risk of bias of included case-control studies and cohort studies using the Newcastle-Ottawa Scale (NOS) (18). Any discrepancies were addressed by consensus. NOS assess the quality of studies in three domains: selection, comparability, and exposure for casecontrol studies; and selection, comparability, and outcome for cohort studies. A maximum of nine stars can be received. Studies with nine stars on the NOS were judged to be at low risk of bias.

Data synthesis and analysis
The primary outcome was VTE. VTE comprises DVT and PE. For studies that reported only DVT or PE, data on DVT or PE was regarded as VTE for analysis. We also conducted analysis on PE. In order to explore the sources of heterogeneity, we conducted several subgroup analyses according to type of antidepressant (selective serotonin reuptake inhibitor (SSRI) vs. tricyclic antidepressant (TCA)), quality of study methodology (low risk of bias vs. high risk of bias) and type of study design (casecontrol studies vs. cohort studies). Sensitivity analyses were conducted by limiting meta-analysis to studies that only included depressed patients or studies that only included female patients. Sensitivity analysis that only included studies that have eliminated the effect of antipsychotics was also conducted.
When possible, adjusted estimates were used; otherwise, unadjusted estimates were calculated with raw data. Random effects model was used, as this takes into account any differences between studies even if there is no statistically significant heterogeneity. We assumed similarity between OR, RR, and HR because VTE and PE were rare events. Publication bias was assessed using Begg's funnel plot and Egger's test (19,20). All statistical analyses were conducted using Stata 12.0.
The Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach was used to assess the quality of evidence (21).

Literature search
Of a total of 1,531 records that were retrieved, 46 were duplicate records. After screening of the titles and abstracts, 1,428 records were excluded. After full text review of the remaining 57 records, 9 studies, including 6 case-control studies involving 40,831 patients (14,15,(22)(23)(24)(25), and 3 cohort studies involving 900,562 patients (13,16,26), were included in the final analysis ( Figure 1). Table 1 shows the characteristics of all the included studies. Of the 6 case-control studies, 2 were nested case-control studies, 2 were population based casecontrol studies, 1 was national case-control study and 1 was hospital based case-control study. Of the 3 cohort studies, 2 were retrospective cohort studies and 1 was prospective cohort study. Of these 9 included studies, 1 case-control study was designed to examine the relationship between corticosteroids exposure and VTE, while data on antidepressant use was also reported (24). One study which aimed to investigate the relationship between depression and VTE, evaluated the association between antidepressant and VTE in depressed patients (depressed patients with antidepressant vs. depressed patients without antidepressant) (16). One study only included female patients (13). The outcome measures were different among studies. Six studies reported data on VTE (13)(14)(15)(16)22,26), 3 studies reported data on DVT (13,25,26), and 3 studies reported data on PE (in 1 of these 3 studies, the outcome measure was fatal PE) (13,23,24). All the 9 studies reported adjusted estimate.

Characteristics of included studies
Of the included studies, only 2 studies (1 casecontrol study (14) and 1 cohort study (26)) received nine stars on the NOS, indicating low risk of bias ( Table 2 and Table 3). For case-control studies, stars were lost in comparability (n = 4), selection of controls (n = 3), definition of the cases (n = 1), ascertainment of exposure (n = 1) ( Table 2). For cohort studies, stars were lost in comparability (n = 1) and ascertainment of exposure (n = 1) ( Table 3). Figure 2 shows the association between antidepressant use and the risk of VTE of all included studies. Overall, the use of antidepressant significantly increased the risk of VTE (9 studies; OR 1.27, 95% CI 1.09 to 1.49; I 2 = 77.1%). When we only included studies reported PE risk in the meta-analysis, the results shows that antidepressant use was associated with an increased risk of PE (3 studies; OR 1.50, 95% CI 1.33 to 1.69; I 2 = 15.9%; Figure 3). In the subgroup analysis by quality of study methodology, a significant increased VTE risk with antidepressant use was observed in studies with high risk of bias (7 studies; OR 1.30, 95% CI 1.12 to 1.51; I 2 = 54.9%; Figure 2), while no association observed in studies with low risk of bias (2 studies; OR 1.27, 95% CI 0.84 to 1.92; I 2 = 91.1%; Figure 2). 1 study reported the relationship between antidepressant use and VTE in depressed patients. The study showed there was no significant Studies included in meta-analysis (n = 9) difference in the VTE risk between depressed patients with and without antidepressant use (OR 0.72, 95% CI 0.39 to 1.31; Table 4). On the contrary, other studies showed that antidepressant use was associated with an increased VTE risk (8 studies; OR 1.31, 95% CI 1.12 to 1.53; I 2 = 77.9%). Table 4 summarizes the results of subgroup analyses and sensitivity analyses. There was no publication bias (Begg's test P = 1.000, Egger's test P = 0.576). With GRADE approach, the quality of evidence was rated as very low because of risk of bias and inconsistency ( Table 5).

SSRI and risk of VTE
Five studies evaluated the association between SSRI use and the risk of VTE. The results showed there was no significant difference in the VTE risk between patients with and without SSRI use (OR 1.10, 95% CI 0.90 to 1.34; I 2 = 62.3%; Figure 4).
Subgroup analysis by quality of study methodology did not change the results (studies with low risk of bias: OR 1.05, 95% CI 0.94 to 1.18; I 2 = 0.0%; studies with high risk of bias: OR 1.05, 95% CI 0.71 to 1.55; I 2 = 73.5%; Figure 4). No significant difference in the VTE risk between patients with and without SSRI was observed when we excluded the study with depressed patients (4 studies; OR 1.14, 95% CI 0.93 to 1.40; I 2 = 66.8%). The study with depressed patients reported similar result (OR 0.75, 95% CI 0.40 to 1.38; Table 4). There was no publication bias (Begg's test P = 1.000, Egger's test P = 0.799). With GRADE approach, the quality of evidence was rated as very low because of risk of bias and inconsistency (Table 5).      Figure  5). In the subgroup analysis, a significant increased VTE risk with TCA exposure was observed in studies with high risk of bias (2 studies; OR 1.34, 95% CI 1.17 to 1.54; I 2 = 0.0%; Figure 5), while no association observed in studies with low risk of bias (2 studies; OR 1.20, 95% CI 0.77 to 1.89; I 2 = 84.3%; Figure 5). The study with depressed patients did not report the association between TCA use and VTE. There was no publication bias (Begg's test P = 0.734, Egger's test P = 0.175). With GRADE approach, the quality of evidence was rated as very low because of risk of bias and inconsistency (Table 5).

DISCUSSION
Our meta-analysis indicated that antidepressant use may be associated with an increased risk of VTE; however, no association was observed when we only included studies with low risk of bias and when we only included study with depressed patients. No association between SSRI use and VTE risk was detected in the overall analysis and in the subgroup analysis by quality of study methodology. A significant increased VTE risk with TCA use was observed, and the quality of evidence was rated as very low. However, no association between TCA use and VTE risk was observed when we only included studies with low risk of bias and when we only included studies that have eliminate the effect of antipsychotics. A significant increased VTE risk was detected with antidepressant use, and the quality of evidence was rated as very low by GRADE approach. However, no association was detected when we only included studies with low risk of bias. Such phenomenon was also observed in the TCA group. Only 2 studies were judged to be at low risk of bias. When we pooled the data from these 2 studies with low risk of bias, high heterogeneity was detected (antidepressant: I 2 = 91.1%; TCA: I 2 = 84.3%), which lead the quality of evidence was rated down by GRADE approach. The heterogeneity can be partly explained by the different type of study design. Of these 2 studies with low risk of bias, 1 was casecontrol study and 1 was cohort study. Compared with cohort studies, case-control studies may reflect relatively short-term effects of antidepressant exposure (27). In addition, the duration of antidepressant use may also different between casecontrol studies and cohort studies.  Several previous studies observed a significant increased risk of VTE in depressed patients (28)(29)(30); however, it is unclear whether depression, antidepressant use or even another related factor drove the increased VTE risk (31). In our metaanalysis, an increased VTE risk with antidepressant use was detected; whereas, sensitivity analysis that only included study with depressed patients failed to detect an association between antidepressant use and VTE risk. Because most of the antidepressant drugs were prescribed for depression treatment and most patients without antidepressant use were nondepressed patients, these data seem to indicate that increased VTE risk may actually be associated with the depression itself but not the antidepressant use.
Interestingly, in the subgroup analyses of our study, we found that the relationship between antidepressant use and VTE risk varied according to the type of antidepressant exposure. Increased VTE risk was detected in TCA use group, but not in SSRI use group. The data did not support the previous hypothesis that depression lead to the increased VTE risk. If it is the depression drove the increased VTE risk, the VTE risk would not be changed according to the type of antidepressant. Serotonin is a weak platelet agonist and serotonin potentiates platelet stimulation induced by adenosine diphosphate or thrombin (32). SSRI decrease serotonin level in platelets by inhibiting serotonin reuptake; therefore, SSRI may relate to decreased platelet activation and prolongation of bleeding time (32). This view may explain why SSRI use did not associate with an increased VTE risk. In fact, a systematic review suggested that SSRI may be associated with increased perioperative bleeding events (33). Systematic reviews showed that antipsychotic drugs may be associated with an increased risk of VTE (8,9). As for the first-generation antipsychotics, the OR was 1.74, 95 % CI 1.28-2.37 (8) and 1.72, 95 % CI 1.31-2.24 (9) in these two systematic reviews, respectively. The chemical structure of the TCA is similar to that of the phenothiazines, one kind of the first-generation antipsychotics (14). This view may explain why TCA use increased VTE risk.
Although another meta-analysis examined the association between antidepressant use and the risk of VTE has been published (34), the findings from our manuscript differ from that paper. The previous meta-analysis showed that both TCA and SSRI were associated with an increased VTE risk, however no association between SSRI use and VTE risk was detected in our manuscript. The previous metaanalysis only included 6 studies compared patients with antidepressant versus individuals without antidepressant, so 1 study with data about SSRI was not included into analysis. In addition, fix effects model was used in the previous meta-analysis, although high heterogeneity was found.
The strengths of this study included the comprehensive literature search, critical appraisal of the included studies, planned subgroup analyses and sensitivity analyses, and the use of GRADE approach. However, our meta-analysis has several limitations. First, only observational studies were included in our meta-analysis. No randomized control trial addressing antidepressant use and its effect on VTE have been published. VTE is associated with a lot of risk factors. Although the included studies attempted to minimize the effect of other risk factors, possibly none of them could fully adjust for all these risk factors. This is the reason why we did not conducted sensitivity analysis on idiopathic VTE. Second, the heterogeneity is relatively high in our meta-analysis. There are many differences between the included studies, such as the definition of antidepressant exposure, age and gender of the patients, and so on. We largely divided antidepressants into TCA and SSRI; however, some studies had different categories (14,15). Third, only 9 studies were included and only 2 studies were judged to be at low risk of bias. The number of studies on this topic limited the possibility of further analysis. More high-quality studies examining the relationship between antidepressant use and VTE risk are needed.

CONCLUSIONS
In conclusion, our meta-analysis suggested that there was no association between selective serotonin reuptake inhibitor use and venous thromboembolism risk in the overall analysis and in subgroup analysis of studies with low risk of bias. Tricyclic antidepressant may be associated with an increased venous thromboembolism risk, but the quality of evidence was rated as very low. However, no association between tricyclic antidepressant use and venous thromboembolism risk was observed when we only included studies with low risk of bias.

CONFLICTS OF INTEREST
The author(s) declared no conflicts of interest. One-to-one interview. If available, charts and medical records were reviewed. For outpatients, exposure was defined as current use of drugs at admission. All drugs recorded had to be taken at admission for more than 1 week. Drugs taken prior to admission but discontinued more than 1 week before admission were not recorded. For inpatients, all drugs taken at the time of inclusion were recorded, including drugs prescribed in hospital. Cox proportional hazards regression model. Adjusted with age, sex, current residing within a long-term care facility, recent prior hospitalization, newly diagnosed cancer or concurrent prescription of lithium, estrogen, aspirin or warfarin. MI = myocardial infarction; RF = renal failure; DM = diabetes mellitus; VTE = venous thromboembolism; BMI = body mass index; EDITH = Etude des De´terminants et Interaction de la THrombose veineuse; PE = pulmonary embolism; SSRI = selective serotonin reuptake inhibitor; AF = atrial fibrillation; CVA = cerebral vascular accident; HF = heart failure; DVT = deep vein thrombosis; BP = blood pressure.  (14) * * * * ** * * * A study can receive a maximum of one star for each item within the Selection and Exposure categories and two stars for Comparability.