Differential effects of antidepressant subgroups on risk of acute myocardial infarction: A nested case–control study

Aims

The primary objective of this study was to investigate the association between antidepressants use and the risk of acute myocardial infarction (AMI).

Methods

We conducted a nested case–control study using a primary care database over the period 2002–2015. From a cohort of patients aged 40–99 years, we identified incident AMI cases and randomly selected 5 controls per case, matched to cases for exact age, sex and index date. Exposure to antidepressants were categorised as current, recent, past and nonusers. Adjusted odds ratio (AOR) and 95% confidence interval (CI) were computed using conditional logistic regression to assess the association between the current use of different antidepressants subgroups and AMI as compared to nonuse. Dose and duration effects were explored.

Results

Totals of 24 155 incident AMI cases and 120 775 controls were included. The current use of antidepressants as a group was associated with a reduced risk (AOR = 0.86; 95% CI: 0.81–0.91), but mainly driven by selective serotonin reuptake inhibitors (AOR = 0.86; 95% CI:0.81–0.93). A reduced risk was also observed with trazodone (AOR = 0.76;95% CI: 0.64–0.91), and clomipramine (AOR = 0.62; 95% CI: 0.40–0.96), whereas no significant effect was observed with other antidepressants. A duration‐dependent effect was suggested for selective serotonin reuptake inhibitors, trazodone and clomipramine, while there was no clear dose‐dependency.

Conclusion

This study suggests that current use of antidepressants interfering selectively with the reuptake of serotonin, and those antagonizing the 5‐HT_2A receptor, are associated with a decrease in AMI risk and should be the antidepressants of choice in patients at cardiovascular risk.

1.

What is already known about this subject

• Several studies have analysed the association of antidepressants with acute myocardial infarction with conflicting results.

• Serotonin plays a role in promoting the platelet aggregation.

• Antidepressants vary considerably in their mechanism of action and, consequently, they may show a different effect on the atherothrombotic risk.

What this study adds

• Selective serotonin reuptake inhibitors, clomipramine (a tricyclic antidepressant with a prominent selectivity on serotonin transporter) and trazodone (an antagonist of 5‐HT_2A receptor, which mediates the action of serotonin in platelets) were associated with a decreased risk of acute myocardial infarction, while other antidepressants showed no benefit.

• The favourable cardiovascular profile shown by these drugs should be considered by the clinicians at the time of selecting an antidepressant in patients with cardiovascular risk.

2. INTRODUCTION

Antidepressants are among the most frequently prescribed drugs worldwide. ¹ , ² Their use is increasing not only for depression but also for a variety of conditions. ² , ³ Generally, all antidepressants show a similar therapeutic efficacy, and the choice of an antidepressant is often based on its mechanism of action, onset of action and its side‐effects profile. ³ Some of their side effects relate to the cardiovascular system ⁴ and thus, the selection of an antidepressant with an adequate benefit–risk profile is crucial, particularly in patients at high cardiovascular risk.

In the last 2 decades, several studies have analysed the association of antidepressants with acute myocardial infarction (AMI), but the results have been inconsistent, some studies reporting a reduced risk, ⁵ , ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ while others suggesting an increased risk. ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ The explanation for such conflicting results is elusive, but it may be related to 2 different factors. On one hand, depression has been reported to increase the risk of AMI, ²¹ and a confounding by indication is expected with all antidepressants. On the other hand, the potential protective effect may not be shared by all drugs, but only for those showing a relevant serotonin‐mediated antiplatelet effect by either depleting the serotonin (also referred to as 5‐hydroxytryptamine; 5‐HT) content in platelets, as a consequence of blocking its reuptake mechanism, ²² or by antagonizing the specific 5‐HT_2A receptor, abundantly present in the platelet membrane and known to mediate the activation of platelets by serotonin. ²³ The inhibition of one of these mechanisms would reduce the ability of platelets to aggregate in an effective way and it may help to reduce the risk of atherothrombotic events. We postulate that the use of selective serotonin reuptake inhibitors (SSRIs) and 5‐HT_2A antagonists (e.g. trazodone), is associated with a decreased risk of AMI, while antidepressants that do not interfere with these serotonin‐mediated mechanisms would have a null effect. We conducted this study with the aim to test this hypothesis, to explore the dose and duration dependence and to identify potential effect modifiers.

3. PATIENTS AND METHODS

3.1. Study population and data source

Data used for this study were derived from Base de datos para la Investigación Farmacoepidemiológica en Atención Primaria (BIFAP), an electronic primary care database validated for pharmacoepidemiological research. ²⁴ The database is updated each year and for this study we used the version updated in 2016, which included anonymised clinical records from 8 million patients belonging to 9 autonomous regions (out of 17), representing around 17% of the total Spanish population and 58% of the population from contributing regions. ²⁴

The electronic clinical records contain anonymised demographic information, prescription details (including the exact date of prescription), clinical events, specialist referrals and results from laboratory and other exploratory tests. The medical diagnoses are recorded using the International Classification of Primary Care, version 2 (ICPC‐2) in 8 regions and the International Classification of Diseases, version 9, Clinical Modification (ICD‐9‐MC) in 1 region. This information is often enriched with free text comments from the primary care physician (PCP). Prescription data in BIFAP include product name, quantity dispensed, dosage regimens, strength and indication. These prescriptions are coded according to the Anatomical Therapeutic Chemical classification scheme. ²⁵ The population included in BIFAP is representative of the Spanish population with respect to age and sex. ²⁴ This database has been widely used for pharmacoepidemiological research, ²⁴ , ²⁶ , ²⁷ , ²⁸ and successfully compared with other well‐known European databases ²⁹ , ³⁰

3.2. Study design

We carried out a case–control study nested in a defined primary cohort composed of patients aged 40–99 years, who had at least 1 year of registry with their PCP during the period 1 January 2002–31 December 2015 and had no record of cancer or a previous AMI. The patients who had a record of cancer were excluded from the cohort as cancer may reduce the life expectancy and they usually have hospital‐dispensed polypharmacy not recorded in the database. Once they fulfilled all the above criteria (start date), they were followed up until the earliest occurrence of 1 of the following endpoints: the outcome of interest AMI, age 100 years, a diagnosis of cancer, death or end of the study period, whichever came first.

3.3. Case definition and ascertainment

From the electronic clinical records, all patients with a code for AMI (ICPC‐2 code K75; ICD‐9‐MC code 410), or a text associated with a diagnosis compatible with AMI were identified as potential cases. Then, 2 researchers blind to any drug exposure carried out the case validation which rendered a positive predictive value of 87.2% (95% confidence interval [CI]: 84.1–89.8; supplementary methods). The potential cases were considered valid if they had a recorded diagnosis of AMI and, additionally, supportive information, including a hospital report, the reference to elevated creatine phosphokinase and/or troponin levels, an electrocardiogram with a Q wave, or the undertaking of revascularization procedures. Only incident AMI cases were considered. The index date was defined as the date of the first AMI record.

3.4. Controls

Five controls per case, individually matched to cases on exact age, sex and index date (the index date of a case was applied to its corresponding controls set), were randomly selected from the primary cohort.

3.5. Exposure definition

Patients were defined as current users of a specific antidepressant subgroup when their last prescription lasted until index date or ended within 30 days prior to the index date, recent users when ended between 31 and 365 days before the index date, past users when ended before 365 days prior to the index date and nonusers if there was no prescription of the antidepressant subgroup studied before the index date (reference category). Current users of each subgroup were further divided according to the daily dose (of the last prescription), continuous duration of treatment and individual drugs for those with enough patients exposed. Two categories for daily dose were used: low and intermediate–high dose (see cut‐off levels in Table S1). Continuous duration of treatment was computed by summing up consecutive prescriptions of antidepressants pertaining to the same antidepressant subgroup. Two prescriptions were considered consecutive when the time elapsed between the end of supply of 1 and the start of the next was 90 days or less. Duration was classified in 2 categories: ≤1 year and >1 year.

Antidepressants marketed in Spain were classified according to their mechanism of action in the following subgroups ³¹ :

• SSRIs: citalopram, escitalopram, fluoxetine, fluvoxamine, paroxetine and sertraline.

• Serotonin‐norepinephrine reuptake inhibitors (SNRIs): desvenlafaxine, venlafaxine and duloxetine.

• Tricyclic antidepressants (TCAs) with a preferential inhibitory action on the reuptake of serotonin (SER): imipramine, amitriptyline, trimipramine, dothiepin (also known as dosulepine) and clomipramine. ³² , ³³ Clomipramine shows a selectivity factor closed to SSRIs (K_D = 0.28 nM for SER transporter and K_D = 38 nM for norepinephrine [NE] transporter), ³³ and sometimes is classified in the SSRIs subgroup.

• Serotonin receptor antagonists with serotonin reuptake inhibition (SARIs): Trazodone (antagonist of 5‐HT_2A receptor; at daily doses of 150 mg or higher also inhibits the SER transporter).

• Noradrenergic α‐2 receptor antagonists with specific serotonergic antagonism (NaSSa): mirtazapine, mianserine.

• Other antidepressants: a miscellaneous subgroup sharing the lack of action on serotonergic pathways, including TCAs with a preferential inhibitory action on NE reuptake (nortriptyline, doxepine) ³² , ³³ ; selective NE reuptake inhibitors (maprotiline and reboxetine), ³² monoamine oxidase inhibitors (moclobemide, tranylcypromine), and melatonergic agonists (agometaline).

3.6. Data analysis

The potential association between the use of each specific antidepressants subgroup and the risk of AMI was assessed by computing the odds ratio (ORs) and its 95% CIs among current users of antidepressants of interest as compared to nonusers through conditional logistic regression. All antidepressant subgroups were included in the model simultaneously. ORs were adjusted (AOR) for possible confounding factors (measured at index date) selected by expert knowledge and included simultaneously in the model, avoiding data‐driven methods such as stepwise regression. The factors included history of cerebrovascular events (transient ischaemic accident, haemorrhagic stroke, ischaemic stroke and unspecified stroke), angina pectoris (defined as such or by nitrate use), peripheral artery disease, hypertension, diabetes mellitus (defined as such or by use of glucose lowering drugs), dyslipidaemia (defined as such or by use of lipid lowering drugs), rheumatoid arthritis, and chronic renal failure. The current use of the following comedications was also included as potential confounders: antiplatelet drugs, oral anticoagulants, nonsteroidal anti‐inflammatory drugs, corticosteroids, angiotensin converting enzyme inhibitors, angiotensin II receptor blockers, calcium channel blockers, β‐blockers, α‐blockers, diuretic drugs, proton‐pump inhibitors and H2‐antihistamines. The full model also included the number of PCP visits in the year prior to the index date as a marker of comorbidity, smoking, alcohol abuse (as recorded by the PCP) and body mass index (BMI, expressed in kg/m²). When the association with a specific pharmacological subgroup was analysed, we adjusted for the use of the other subgroups.

The missing values of BMI and smoking were included in specific categories. This missing‐indicator method was considered appropriate as we checked that the distribution of missing values did not vary across main exposure categories. ³⁴ However, in a sensitivity analysis we also run models based in multiple imputation by chained equations (MICE). ³⁵

We assessed the potential interaction of antidepressants use with age (<70 and ≥70 years), sex and background cardiovascular risk (high risk: patients with a history of angina pectoris, peripheral arterial disease, stroke or diabetes; intermediate risk: those with any risk cardiovascular risk factor, including hypertension, dyslipidaemia, chronic renal failure, smoking and BMI > 30 kg/m²; and low risk: the remainder). The AORs across strata were compared using the test of interaction described by Altman and Bland. ³⁶

The analysis was performed using STATA software, version 15/SE (StataCorp. College Station, Texas. 77845, USA).

3.7. Sensitivity analyses

We performed 3 sensitivity analyses: (i) new‐user analysis, in which we excluded from both cases and controls all patients who filled an antidepressant prescription any time before the start date of follow up (prevalent users) ³⁷ ; (ii) we run the model adding antecedents of depression and assessed whether this variable had an impact on the association with the different antidepressants examined; and (iii) we performed MICE analysis to address the missing values for smoking and BMI.

3.8. Ethical review

Access to anonymised data from BIFAP was granted by the BIFAP Scientific Committee (project #04/2016; approval date: 26 May 2016). According to the Spanish law no specific ethical review was required for studies using fully anonymised data.

4. RESULTS

The study included 24 155 AMI cases and 120 775 matched controls (Figure 1). The mean follow‐up from since start date to index date was 3.76 (standard deviation [SD] ± 3.00) years for cases and 3.61 (± 2.90) years for controls. As expected, cases presented a higher prevalence of cardiovascular comorbidities and cardiovascular drugs than controls (Table 1). As a complement, we show in Table S2 and Figure S1 the distribution of these factors in current users of different antidepressants subgroups as compared to nonusers among controls (a representative sample of the source population). Broadly, the cardiovascular comorbidity was higher among users of antidepressants than among nonusers, emphasizing the importance of full adjustment.

FIGURE 1.

Flowchart of case selection

TABLE 1.

Characteristics of cases and controls.

	Cases (%) n = 24 155	Controls (%) n = 120 775	Crude OR (95% CI) †
Men	17 208 (71.2)	86 040 (71.2)	–
Age (y), mean (±SD)	67.1 (± 13.4)	67.0 (± 13.4)	–
Number of visits
<6	6867 (28.4)	44 979 (37.2)	1 (reference)
6–15	9008 (37.3)	43 706 (36.2)	1.44 (1.39–1.49)
15–24	4498 (18.6)	18 381 (15.2)	1.81 (1.73–1.89)
>24	3782 (15.7)	13 709 (11.4)	2.12 (2.02–2.23)
BMI, kg/m2
<25	2721 (11.3)	14 574 (12.1)	1 (reference)
25–29	6997 (29.0)	34 041 (28.2)	1.10 (1.05–1.16)
30–34	4162 (17.2)	18 752 (15.5)	1.19 (1.13–1.26)
35–49	1122 (4.7)	4488 (3.7)	1.35 (1.25–1.46)
≥40	334 (1.38)	1149 (0.95)	1.56 (1.38–1.78)
Unknown	8819 (36.5)	47 771 (39.6)	0.98 (0.94–1.03)
Smoking
Never smoking	5489 (22.7)	32176 (26.6)	1 (reference)
Current smoker	6498 (26.9)	20135 (16.7)	2.04 (1.95–2.12)
Past smoker	1287 (5.3)	7067 (5.9)	1.11 (1.04–1.19)
Unknown	10 881 (45.1)	61 397 (50.8)	1.07 (1.03–1.11)
Alcohol abuse	655 (2.7)	3011 (2.5)	1.09 (1.00–1.19)
Cerebrovascular event
Ischaemic	600 (2.5)	2204 (1.82)	1.39 (1.27–1.53)
Haemorrhagic	89 (0.37)	354 (0.29)	1.28 (1.02–1.62)
Unspecified	427 (1.77)	1818 (1.51)	1.20 (1.08–1.34)
TIA	503 (2.1)	2015 (1.67)	1.28 (1.16–1.42)
Angina pectoris	2735 (11.3)	5259 (4.4)	2.91 (2.77–3.06)
PAD	1092 (4.5)	2478 (2.1)	2.30 (2.14–2.48)
Hypertension	12 534 (51.9)	52 407 (43.4)	1.49 (1.45–1.53)
Diabetes	6543 (27.1)	19 966 (16.5)	1.92 (1.85–1.98)
Dyslipidaemia	11 355 (47.0)	42 725 (35.4)	1.67 (1.62–1.72)
Rheumatoid arthritis	238 (0.99)	757 (0.63)	1.58 (1.37–1.83)
Chronic renal failure	919 (3.8)	2919 (2.4)	1.62 (1.50–1.75)
Comedication
Antiplatelets	4793 (19.8)	14 652 (12.1)	2.05 (1.97–2.13)
Anticoagulants	921 (3.8)	5018 (4.2)	0.92 (0.85–0.99)
Beta blockers	2666 (11.0)	7654 (6.3)	1.91 (1.83–2.01)
Alpha blockers	609 (2.5)	2497 (2.1)	1.24 (1.13–1.36)
ACEIs	4212 (17.4)	17 351 (14.4)	1.37 (1.32–1.42)
ARBs	3773 (15.6)	14 683 (12.2)	1.43 (1.37–1.48)
CCBs	3316 (13.7)	11 488 (9.5)	1.63 (1.56–1.70)
Diuretics	2825 (11.7)	10 717 (8.9)	1.47 (1.40–1.54)
NSAIDs	2440 (10.1)	11 103 (9.2)	1.19 (1.13–1.25)
Corticosteroids	523 (2.2)	1798 (1.49)	1.50 (1.35–1.65)
PPIs	6494 (26.9)	25 538 (21.2)	1.53 (1.47–1.59)
H2–receptor antagonists	521 (2.2)	1661 (1.38)	1.61 (1.46–1.78)

^†Adjusted only for matching factors (age, sex and index date).

Abbreviations: OR: odds ratio; CI: confidence interval; SD: standard deviation; BMI: body mass index; CVE: cerebrovascular event; TIA: transient ischemic attack; AMI: acute myocardial infarction; PAD: peripheral artery disease; ACEIs: angiotensin converting enzyme inhibitors; ARBs: angiotensin II receptor blockers; CCBs: calcium channel blocker; NSAIDs: nonsteroidal anti‐inflammatory drugs; PPIs: proton‐pump inhibitors.

Note: Percentages ≥2 have been rounded to the first decimal place. ORs and percentages <2 have been rounded to the second decimal.

4.1. Association between AMI and antidepressant drugs as a group

Among cases, 1912 were current users (7.92%) of antidepressants and 9283 (7.69%) among controls, yielding an AOR = 0.86 (95% CI: 0.81–0.91). Such a reduced risk disappeared upon discontinuation (AOR = 1.06; 95% CI: 0.98–1.14) and was more evident when treatment lasted >1 year (AOR_{>1 year} = 0.83; 95% CI: 0.78–0.89; AOR_{≤1 year} = 0.92; 95% CI: 0.84–1.00).

4.2. Risk of AMI by antidepressant subgroups and individual drugs

We observed a reduced risk of AMI among current users of SSRIs (AOR = 0.86; 95% CI: 0.81–0.93), and current users of SARIs (AOR = 0.76; 95% CI: 0.64–0.91), which disappeared in both upon discontinuation (Table 2). The results suggest a trend for a duration‐dependent effect in both subgroups (Table 3). For TCAs‐SER, a trend for a duration‐dependent effect was also suggested (AOR_{>1 year} = 0.77; 95% CI: 0.62–0.96), mostly driven by clomipramine (the exclusion of this drug from the subgroup rendered the AOR nonsignificant (0.86; 95% CI: 0.67–1.08). No clear dose‐dependency was found with any subgroup (Table S3).

TABLE 2.

Risk of acute myocardial infarction associated with the use of different antidepressant subgroups.

	Cases (%) n = 24 155	Controls (%) n = 120 775	Crude OR (95% CI) †	Adjusted OR (95% CI) ‡
Selective serotonin reuptake inhibitors
Nonusers	21 044 (87.1)	106 707 (88.4)	1 (reference)	1 (reference)
Current	1150 (4.8)	5659 (4.7)	1.04 (0.97–1.11)	0.86 (0.81–0.93)
Recent	684 (2.8)	2595 (2.2)	1.34 (1.23–1.46)	1.12 (1.02–1.23)
Past	1277 (5.3)	5814 (4.8)	1.12 (1.05–1.19)	0.96 (0.90–1.03)
Serotonin–norepinephrine reuptake inhibitors
Nonusers	23 401 (96.9)	117 480 (97.3)	1 (reference)	1 (reference)
Current	319 (1.32)	1452 (1.20)	1.11 (0.98–1.25)	0.93 (0.82–1.06)
Recent	131 (0.54)	598 (0.50)	1.10 (0.91–1.33)	0.90 (0.74–1.10)
Past	304 (1.26)	1245 (1.03)	1.23 (1.08–1.39)	1.07 (0.93–1.22)
Tricyclic antidepressants with a preferential inhibitory action on the reuptake of serotonin
Nonusers	23 269 (96.3)	117 195 (97.0)	1 (reference)	1 (reference)
Current	199 (0.82)	923 (0.76)	1.09 (0.93–1.27)	0.92 (0.79–1.08)
Recent	162 (0.67)	650 (0.54)	1.26 (1.06–1.50)	1.09 (0.91–1.31)
Past	525 (2.2)	2007 (1.66)	1.32 (1.20–1.46)	1.16 (1.04–1.29)
Serotonin receptor antagonists with serotonin reuptake inhibition
Nonusers	23 757 (98.4)	118 811 (98.4)	1 (reference)	1 (reference)
Current	154 (0.64)	845 (0.70)	0.91 (0.77–1.08)	0.76 (0.64–0.91)
Recent	80 (0.33)	420 (0.35)	0.95 (0.75–1.21)	0.81 (0.63–1.03)
Past	164 (0.68)	699 (0.58)	1.17 (0.99–1.39)	1.00 (0.83–1.20)
Noradrenergic α‐2 receptor antagonists with specific serotonergic antagonism
Nonusers	23 449 (97.1)	117 479 (97.3)	1 (reference)	1 (reference)
Current	274 (1.13)	1315 (1.09)	1.04 (0.92–1.19)	0.92 (0.80–1.05)
Recent	130 (0.54)	581 (0.48)	1.12 (0.93–1.36)	0.97 (0.79–1.18)
Past	302 (1.25)	1400 (1.16)	1.08 (0.95–1.23)	0.96 (0.84–1.10)
Other antidepressants
Nonusers	24 000 (99.4)	120 029 (99.4)	1 (reference)	1 (reference)
Current	46 (0.19)	229 (0.19)	1.00 (0.73–1.38)	0.93 (0.67–1.29)
Recent	36 (0.15)	144 (0.12)	1.25 (0.87–1.80)	1.10 (0.75–1.61)
Past	73 (0.30)	373 (0.31)	0.98 (0.76–1.26)	0.84 (0.65–1.10)

^†Adjusted for matching factors (age, sex and index date).

^‡Adjusted for matching factors (age, sex and index date), the variables included in Table 1 and all pharmacological subgroups, except that evaluated.

Note: Percentages ≥2 have been rounded to the first decimal place. ORs and percentages <2 have been rounded to the second decimal.

TABLE 3.

Risk of acute myocardial infarction associated with the use of different antidepressant subgroups by duration of treatment among current users as compared to nonusers.

	Cases (%) n = 24 155	Controls (%) n = 120 775	Crude OR (95% CI) †	Adjusted OR (95% CI) ‡
Selective serotonin reuptake inhibitors
≤1 year	464 (1.92)	2193 (1.82)	1.08 (0.97–1.19)	0.91 (0.81–1.01)
>1 year	686 (2.84)	3466 (2.87)	1.01 (0.93–1.10)	0.84 (0.77–0.91)
Serotonin–norepinephrine reuptake inhibitors
≤1 year	117 (0.48)	529 (0.44)	1.11 (0.91–1.36)	0.95 (0.77–1.18)
>1 year	202 (0.84)	923 (0.76)	1.10 (0.94–1.28)	0.92 (0.78–1.08)
Tricyclic antidepressants with a preferential inhibitory action on the reuptake of serotonin
≤1 year	96 (0.40)	360 (0.30)	1.35 (1.07–1.69)	1.15 (0.91–1.46)
>1 year	103 (0.43)	563 (0.47)	0.92 (0.75–1.14)	0.77 (0.62–0.96)
Serotonin receptor antagonists with serotonin reuptake inhibition
≤1 year	88 (0.36)	447 (0.37)	0.98 (0.78–1.24)	0.79 (0.62–1.01)
>1 year	66 (0.27)	398 (0.33)	0.83 (0.64–1.08)	0.73 (0.56–0.96)
Noradrenergic α‐2 receptor antagonists with specific serotonergic antagonism
≤1 year	121 (0.50)	550 (0.46)	1.10 (0.90–1.34)	0.97 (0.79–1.19)
>1 year	153 (0.63)	765 (0.63)	1.00 (0.84–1.20)	0.88 (0.74–1.06)
Other antidepressants
≤1 year	13 (0.05)	100 (0.08)	0.65 (0.36–1.16)	0.61 (0.34–1.10)
>1 year	33 (0.14)	129 (0.11)	1.28 (0.87–1.87)	1.19 (0.80–1.77)

^†Adjusted for matching factors (age, sex and index date).

^‡Adjusted for matching factors (age, sex and index date), the variables included in Table 1 and all pharmacological subgroups, except that evaluated.

The AORs of AMI associated with the current use of individual antidepressant drugs (with at least 10 exposed cases) appear in Table 4. Escitalopram (AOR = 0.80; 95% CI: 0.70–0.92), sertraline (AOR = 0.83; 95% CI: 0.72–0.97), clomipramine (AOR = 0.62; 95% CI:.40–0.96) and trazodone (AOR = 0.76; 95% CI: 0.64–0.91) showed the greatest effects.

TABLE 4.

Risk of acute myocardial infarction associated with the current use of individual antidepressant drugs ^† as compared to nonusers.

Subgroup ‡ and individual agent	Cases (%) n = 24 155	Controls (%) n = 120 775	Crude OR (95% CI) §	Adjusted OR (95% CI)
Selective serotonin reuptake inhibitors
Citalopram	199 (0.82)	1000 (0.83)	1.00 (0.86–1.16)	0.86 (0.73–1.00)
Escitalopram	268 (1.11)	1426 (1.18)	0.94 (0.83–1.08)	0.80 (0.70–0.92)
Fluoxetine	173 (0.72)	685 (0.57)	1.27 (1.08–1.51)	1.06 (0.89–1.27)
Fluvoxamine	14 (0.06)	94 (0.08)	0.74 (0.42–1.31)	0.68 (0.38–1.22)
Paroxetine	296 (1.23)	1429 (1.18)	1.04 (0.92–1.18)	0.88 (0.77–1.00)
Sertraline	217 (0.90)	1111 (0.92)	0.98 (0.85–1.13)	0.83 (0.72–0.97)
Serotonin–norepinephrine reuptake inhibitors
Venlafaxine	203 (0.84)	920 (0.76)	1.11 (0.95–1.29)	0.95 (0.81–1.12)
Duloxetine	117 (0.48)	506 (0.42)	1.16 (0.95–1.42)	0.97 (0.78–1.20)
Tricyclic antidepressants with a preferential inhibitory action on the reuptake of serotonin
Imipramine	12 (0.05)	34 (0.03)	1.76 (0.91–3.41)	1.57 (0.79–3.11)
Amitriptyline	161 (0.67)	710 (0.59)	1.14 (0.96–1.36)	0.95 (0.80–1.14)
Clomipramine	24 (0.10)	174 (0.14)	0.69 (0.45–1.06)	0.62 (0.40–0.96)
Serotonin receptor antagonists with serotonin reuptake inhibition
Trazodone	154 (0.64)	845 (0.70)	0.91 (0.77–1.08)	0.76 (0.64–0.91)
Noradrenergic α‐2 receptor antagonists with specific serotonergic antagonism
Mirtazapine	225 (0.93)	1092 (0.90)	1.03 (0.89–1.19)	0.92 (0.79–1.07)
Mianserine	51 (0.21)	227 (0.19)	1.12 (0.83–1.52)	1.00 (0.73–1.37)

^†With at least 10 exposed cases

^‡The subgroup Other antidepressants had very few patients exposed to individual drugs and is not shown.

^§Adjusted for matching factors (age, sex and index date).

¶Adjusted for matching factors (age, sex and index date), the variables included in Table 1 and all antidepressants, except the 1 evaluated.

Note: Percentages ≥2 have been rounded to the first decimal place. ORs and percentages <2 have been rounded to the second decimal.

Note: Percentages ≥2 have been rounded to the first decimal place.

ORs and percentages <2 have been rounded to the second decimal.

4.3. Interaction with sex, age and cardiovascular risk

Among current users of SSRIs and SARIs we did not find a statistically significant interaction with sex or age. Further, we did not find a statistically significant interaction with background cardiovascular risk (Figure 2), although results suggest that in low‐risk patients SSRIs or SARIs may not show a protective effect, but the numbers are small. With clomipramine we did not have enough sample size to perform a meaningful interaction analysis.

FIGURE 2.

Risk of acute myocardial infarction (AMI) associated with selective serotonin reuptake inhibitors (SSRIs) and serotonin receptor antagonists with serotonin reuptake inhibition (SARIs) by sex, age and background cardiovascular risk. AOR: adjusted odds ratio; CI: confidence interval; CV: cardiovascular

4.4. Sensitivity analyses

The main results did not materially change when the analysis was restricted to new users (Table S4), or when we used MICE models for addressing the missing values of smoking and BMI (Table S5). The inclusion of antecedents of depression as a covariate in the model did not change, either, the estimates for SSRIs (AOR = 0.86; 95% CI: 0.79–0.92) or for any other subgroup (data not shown). In addition, a stratified analysis did not show any difference (AOR for SSRIs among patients with a record of depression = 0.85; 95% CI: 0.70–1.04; AOR for SSRIs in patients without a record of depression = 0.86; 95% CI: 0.77–0.97).

5. DISCUSSION

Our results show that current use of antidepressants was associated with a lower risk of AMI, but such an effect was not uniformly distributed across subgroups and individual antidepressants. We only found evidence of a risk reduction of AMI for SSRIs, trazodone (SARI) and clomipramine, in particular when used for >1 year. Such a reduced risk of AMI was not significantly modified by sex, age or background cardiovascular risk.

Several pharmacoepidemiological studies have investigated the association between antidepressants use and the risk of AMI with conflicting results. ⁵ , ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ Depression has been reported as a risk factor for AMI and then a possible explanation for the reported discrepancies could be a confounding by indication if results were not adjusted by antecedents of depression. ¹⁴ , ¹⁷ , ¹⁸ However, in our study, we did not observe a relevant change when the variable antecedents of depression was or not included in the model. In some studies, the small number of antidepressant users might also be a limitation to show a protective effect. ⁷ , ¹¹ Our results suggest another explanation: the protective effect may only be observed with some antidepressants, depending on their mechanism of action. Two meta‐analyses have been published so far reporting results for SSRIs: Oh et al ³⁸ reported an overall OR of 0.93 (95% CI: 0.65–1.33), including 12 studies and using coronary heart disease as the outcome; and Undela et al ³⁹ reported a pooled OR of 0.84 (95% CI: 0.57–1.22) including 11 studies and using MI as the outcome of interest. They also reported results for TCAs: 1.51 (95% CI: 1.07–2.12; 13 studies) and 1.14 (95% CI: 0.67–1.96; 6 studies), respectively. Three new studies have been published after these meta‐analyses reporting results for SSRIs, 1 showing a significant decreased risk, ¹³ 1 a marginally nonsignificant decreased risk ⁵ and another showing no effect. ¹⁵ Overall, we found 15 published studies focusing on SSRIs and MI: 4 found a decreased risk, 8 no effect (but most of them with wide confidence intervals), and 3 an increased risk (see Table S6).

We found evidence of a reduced risk of AMI only for SSRIs, trazodone and clomipramine, which have in common their interference with serotonin actions that may ultimately affect platelet aggregation. ²² SSRIs are well known to inhibit the reuptake of serotonin in both neurons and platelets, as the serotonin transporter is the same. ²² Through this reuptake mechanism, platelets upload serotonin from plasma in their dense‐granules and release it, along with other substances such as adenosine diphosphate, when they are activated. Serotonin is by itself a weak platelet activator, but its release greatly potentiates the aggregation induced by adenosine diphosphate, epinephrine and collagen, ⁴⁰ creating a positive feedback loop that promotes further platelet activation and vasoconstriction. ⁴¹ After 4–8 weeks of SSRIs use, the serotonin content in platelets is almost depleted ⁴² and, through this mechanism, SSRIs could alter platelet function, which is clinically expressed by a prolongation of bleeding time ⁴³ and an increase in the risk of bleeding, ²² with particularly strong evidence being upper gastrointestinal bleeding. ⁴⁴ , ⁴⁵ The beneficial counterpart of this antiplatelet action could be the reduction of the atherothrombotic risk. Our data strongly support this hypothesis and suggest that the effect is duration‐dependent, which seems logical as the more time the patient had the platelet aggregation inhibited, the greater the protective effect (that is, fewer events are accumulated over time). All individual SSRIs showed a statistically significant decreased risk of AMI with the only exception of fluoxetine and fluvoxamine; however, these drugs were the SSRIs with the lowest exposure during the study period and, then, a chance effect cannot be excluded. In the context of this hypothesis, the lack of effect of serotonin–norepinephrine reuptake inhibitors is not consistent, although a small effect cannot be ruled out.

Our data also show that the use of clomipramine is associated with a reduced risk of AMI. Clomipramine is known to have the highest affinity for 5‐HT transporter of all TCAs (K_D = 0.28 nM for SER transporter and K_D = 38 nM for NE transporter), ³³ showing a depletion of serotonin content from platelets similar to the one observed with SSRIs. ⁴⁶ For this reason, in some studies, clomipramine is classified among the SSRIs. ⁴⁴ , ⁴⁷ With amitriptyline or dothiepin, we did not observe a reduced risk, but these drugs show an affinity for the serotonin transporter lower (K_D = 4.30 and 8.60nM, respectively) than the one of clomipramine (K_D = 0.28 nM). ³³ Additionally, in our study, most patients used amitriptyline at low doses (91% used 50 mg or less) which may preclude to find a relevant effect on platelets.

An important novelty of our study is the reduced risk of AMI found with trazodone (the only drug available within the SARIs subgroup). This drug is known to be multifunctional ⁴⁸ : it is an antagonist of the 5‐HT_2A receptor, but at high doses (150 mg or higher) also inhibits the serotonin transporter acting as an SSRI. Often, however, it is used at low doses (100 mg or lower) for its hypnotic property in the elderly. ⁴⁹ For instance, in our study 71% of patients were using trazodone at low doses, the mean age of users being 77.2 (SD ±12.3) years, as compared to 72.1 years (SD ±12.7) among users of all other antidepressants (P = .001). At low doses, trazodone is mainly an antagonist of 5‐HT_2A and H₁ receptors, which accounts for its hypnotic effect. ⁴⁸ It is interesting, that 5‐HT_2A receptors are abundant in membranes of platelets and their activation, once serotonin in released by activated platelets, contribute to thrombus growth. ⁵⁰ Based on this mechanism, 5‐HT_2A antagonists have been proposed as antithrombotic agents. ⁵¹ In line with this evidence, the antithrombotic effect of trazodone would be somewhat expected. However, as far as we know, this is the first time that a reduced risk of AMI is reported for this drug. It is interesting to note that in 1999 our group reported an increased risk of gastrointestinal bleeding associated with trazodone (similar to that found for SSRIs) and we then suggested the antagonism of 5‐HT_2A receptor as the most probable mechanism. ⁴⁴

Although the improvement of depression could also play a protective effect on AMI, the lack of effect shown by other antidepressants acting through mechanisms not related with serotonin supports the hypothesis that the protective effect of SSRIs, trazodone and clomipramine is mostly explained by a direct action on platelet aggregation.

The main strengths of our study are: (i) it is a large study using real‐world data; (ii) controls were randomly selected from the underlying cohort, which prevents a selection bias; (iii) cases were selected after an exhaustive validation exercise, which assured a high positive predictive value; (iv) the information on drug prescriptions in the database is complete as the PCPs filed them through the computer; and (v) new‐user analysis yielded similar results which minimizes the possibility of a prevalent user bias as an explanation of our results.

The study may have some limitations. Firstly, despite taking into account a large number of covariates, we cannot exclude the possibility of a residual confounding by unmeasured variables such as lifestyle factors, including physical activity, nutritional habits or socioeconomic status. However, it is likely that all these factors in depressed patients would spuriously favour a positive association between AMI risk and use of antidepressants; thus, the adjustment for these unmeasured factors would have led to a stronger protective effect. Secondly, we were not able to ascertain the adherence to treatment of patients, but it is unlikely that this potential misclassification was differential between cases and controls. If such nondifferential misclassification of the exposure were relevant, the results would be biased toward the null hypothesis, ⁵² meaning that the real protection would be stronger. Thirdly, although our study included AMI cases regardless of the outcome (fatal or nonfatal), it is possible that fatal cases occurring outside hospital might not have been included in our study and, then, it is uncertain if the protective effect found with some antidepressants could be extrapolated to them. Fourthly, we did not adjust for other indications of antidepressants different from depression (e.g. anxiety) because psychiatric conditions are not accurately recorded in primary care.

In conclusion, the findings of this study provide evidence that current use of antidepressants interfering selectively with the reuptake of serotonin (SSRIs and the TCAs clomipramine), and those antagonizing 5‐HT_2A receptor (trazodone) are associated with a decrease risk of AMI. The results of this study could have implications for the selection of the appropriate antidepressant, in particular for people at high cardiovascular risk.

COMPETING INTERESTS

There are no competing interests to declare.

CONTRIBUTORS

Wrote manuscript: R.A.F., F.d.A. and F.B. Designed research: F.d.A., S.R.M. and M.G. Data collection: S.R.M., D.G.B., M.G., A.G.L. and F.d.A. Analysed data: R.A.F., S.R.M. and F.d.A. Data interpretation: F.d.A, R.A.F., S.R.M., M.G., A.G.L., D.G.B. and F.B. The results, discussion and conclusions are from the authors and do not necessarily represent the position of the Spanish Agency for Medicines and Medical Devices.

PRINCIPAL INVESTIGATOR

The principal investigator of this project is Prof. Francisco J. de Abajo, who is an author of the paper.

Supporting information

FIGURE S1 Pictorial representation of the comorbidity burden among current users of different antidepressant subgroups (red) as compared to nonusers (grey). Broadly, users of antidepressants presented a higher comorbidity burden (in particular cardiovascular diseases) than nonusers. The use of certain drugs are indicators of comorbidity.

TABLE S1 Classification of antidepressant by pharmacological subgroup and cut‐off levels for daily dose.

TABLE S2 Comorbidity burden among current users of different antidepressant subgroups as compared to nonusers.

TABLE S3 Risk of acute myocardial infarction associated with the use of different antidepressant subgroups by daily dose among current users.

TABLE S4 Risk of acute myocardial infarction associated with the use of different antidepressant subgroups among new users.

TABLE S5 Risk of acute myocardial infarction associated with the use of different antidepressant subgroups. Multiple imputation by chained equations (MICE).

TABLE S6 Characteristics of previous studies of antidepressants (selective serotonin reuptake inhibitors and tricyclic antidepressants) and acute myocardial infarction.

Alqdwah‐Fattouh R, Rodríguez‐Martín S, de Abajo FJ, et al. Differential effects of antidepressant subgroups on risk of acute myocardial infarction: A nested case–control study. Br J Clin Pharmacol. 2020;86:2040–2050. 10.1111/bcp.14299

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.