Serious cardiovascular adverse events with fluoroquinolones versus other antibiotics: A self‐controlled case series analysis

Abstract The objective of this study was to evaluate the association between fluoroquinolone (FQ) use and the occurrence of aortic aneurysm/dissection (AA/AD), acute myocardial infarction (AMI), ventricular arrhythmias (VenA), and all‐cause mortality vs other commonly used antibiotics. We conducted a self‐controlled case series analysis of patients who experienced the outcomes of AA/AD, AMI, and VenA, based on diagnosis codes from emergency department visits and hospitalizations within Veterans Health Administration, and death in FY2014‐FY2018. These Veterans also received outpatient prescriptions for FQs. Conditional Poisson regression models were used to estimate the association between FQs and each of the outcomes vs antibiotics of interest (ie amoxicillin or amoxicillin/clavulanate, azithromycin, doxycycline, cefuroxime or cephalexin, or sulfamethoxazole‐trimethoprim), adjusted for time‐varying covariates. Using a 30‐day risk period after each antibiotic prescription, adjusted incidence rate ratios (aIRRs) for FQs vs each comparator antibiotic were not statistically different for outcomes of VenA or AMI. For AA/AD, incidence was higher during FQ risk periods vs amoxicillin [aIRR 1.50 (95% CI 1.01, 2.25)] and azithromycin [aIRR 2.15 (95% CI 1.27, 3.64)] risk periods. A significantly increased risk of mortality was observed with FQs vs each antibiotic of interest. FQs were associated with an increased risk of AA/AD vs amoxicillin and azithromycin and an increased risk of all‐cause mortality vs multiple antibiotics commonly used for outpatient infections. Although the differences in event rates are small, FQ use should be limited to serious infections without appropriate alternatives.


| INTRODUC TI ON
Fluoroquinolone (FQ) prescribing has been steadily declining in the Veterans Health Administration (VHA) over the past decade. This is likely due to multiple factors, including widespread antimicrobial stewardship 1,2 and increased provider awareness of serious adverse drug reactions, such as tendon rupture, irreversible peripheral neuropathy, hypoglycemia, and most recently, aortic rupture and dissection. 3,4 Although FQ prescribing is decreasing, inappropriate use remains a concern; this includes utilization in patients at increased risk for adverse events.
Following the latest FDA warning in December 2018, the VA Center for Medication Safety, as part of its pharmacovigilance program, performed an active surveillance project to assess the potential association between FQ use and aortic aneurysm/dissection, as well as acute myocardial infarction, ventricular arrhythmias, Achilles tendon rupture, peripheral neuropathy, and 30-day all-cause mortality in Veterans. Using propensity score matching with readily available potential confounders and Cox Proportional Hazards regression, the surveillance project detected a potential signal of an increased risk of aortic aneurysm/ dissection, acute myocardial infarction, and 30-day all-cause mortality with FQs compared to both azithromycin and amoxicillin. Therefore, a more comprehensive and rigorous study was conducted using methods that would adjust for additional potential confounding. The objective of this study was to evaluate the association between FQs and the occurrence of aortic aneurysm/dissection (AA/AD), acute myocardial infarction (AMI), ventricular arrhythmias (VenA), and all-cause mortality vs other commonly used antibiotics employing a self-controlled case series analysis (SCCSA).

| Study design and sample construction
Self-controlled case series analysis is useful when the exposure is transient and the outcome acute. [5][6][7] Patients serve as their own controls, so only cases are included, and there is no need to adjust for time-invariant or fixed confounders (eg sex, race/ethnicity).  (Table S1) [8][9][10][11][12] in the primary or principal position for emergency department visits or hospitalizations, respectively, or death during the study period of fiscal years (FYs) 2014 through 2018 and received oral FQs as outpatients in this same time frame were eligible for inclusion ( Figure 1). Patients who received >42 consecutive days of fluoroquinolones (ie chronic therapy) were excluded. The study time frame started one year after their first inpatient stay or outpatient visit to ensure a full baseline year; this was the index date (earliest date was 10/1/2013). The evaluation ended on the date the patient entered hospice/palliative care, the date of death, or the end of the study period (ie 9/30/2018), whichever was earliest. The

Institutional Review Boards for VA Pharmacy Benefits Management
Services and VA Pittsburgh Healthcare System approved the study.

Key points
• In this self-controlled case series analysis, the incidence of aortic aneurysm/dissection was significantly higher during fluoroquinolone vs amoxicillin and azithromycin risk periods.
• Fluoroquinolones were associated with an increased risk of all-cause mortality vs multiple antibiotics commonly used for outpatient infections.
• A significantly increased risk of acute myocardial infarction or ventricular arrhythmias was not observed with fluoroquinolones vs each comparator antibiotic.

| Outcomes
Patients could enter multiple outcome groups and have the same outcome more than once, except for death. Patients with any of the specified ICD-9/10-CM diagnosis codes, in any position, associated with an inpatient or outpatient visit within one year prior to the index date were excluded to try to identify incident events. For patients with the outcome of death, those with only one observation period (ie periods of time when patient is at risk of an outcome due to receipt of an antibiotic or periods of time when patient is not at risk because no antibiotics were received) were excluded because SCCSA (within person comparisons) could not be conducted ( Figure 1).

| Data sources
Data on demographics, comorbidities, inpatient/outpatient encounters for VenA, AA/AD, and AMI, and common respiratory, urinary, and skin and soft-tissue infections were obtained from the Inpatient

| Construction of observation periods
Observation periods included risk periods and non-risk periods (ie no antibiotic). The antibiotic risk periods were 30 days, 8,13 or entire day-supply (whichever was greater, up to a maximum of 42 days),

| Time-varying and fixed covariates
To describe the patients who experienced each of the outcomes, we collected data on demographics (ie age, sex, race/ethnicity), smoking status, and comorbidities, as defined in the Quan coding algorithm for the Charlson Comorbidity Index (CCI), at baseline. 14 We also pulled data on several comorbidities that are risk factors for the

Study Time Frame
ICD-9/10-CM diagnosis codes, in any position, associated with outpatient or emergency department visits or hospitalizations within seven days before the start of the antibiotic risk period through seven days after (Table S1).

| Statistical analysis
We outcome of the same type. We also examined a 10-day risk period for all outcomes and 60 days for AA/AD as sensitivity analyses. 9,11,12 A two-sided P < .05 was used to indicate statistical significance.
Analyses were performed using SAS version 9.4 (SAS Institute Inc, Cary, NC) and STATA 14 (College Station, TX).

| RE SULTS
The outcome groups included 3154 patients with VenA, 2027 patients with AA/AD, 13 504 patients with AMI, and 109 024 patients who died and received at least one outpatient prescription for a FQ ( Figure 1). Table 1 includes the proportions of patients who received prescriptions for each of the antibiotics, by outcome. For all four outcome groups, patients were predominantly male, and the majority were white (Table 2). At baseline, the mean age of patients was approximately 68 years old for all outcomes, except mortality, which was 72.5 years. The percentage of current smokers was 36%-38% in all outcome groups, except AA/AD, in which it was 53%.
Using a 30-day risk period, the aIRRs for FQs vs each comparator antibiotic of interest for the outcomes of VenA and AMI were not statistically significant (

| D ISCUSS I ON
VHA is one of the largest integrated health systems in the United States, with large databases that provide the ideal mechanism to TA B L E 1 Proportion of patients, by outcome, who received outpatient prescriptions for the comparator antibiotics during the study time frame  study rare, but serious adverse drug reactions that were not identified during pre-marketing trials. Our findings in Veterans suggest an increased incidence of AA/AD with the FQs vs both amoxicillin and azithromycin. We also found an increased incidence of 30-day, allcause mortality with the FQs vs each comparator antibiotic of interest. The SCCSA automatically controls for both known and unknown time-invariant confounders as patients serve as their own controls.
In addition, this removes any bias that may be introduced in the selection of controls. Finally, we also included potential time-varying confounders in the model to try to limit residual confounding due to differences between patients who receive FQs vs other antibiotics.
Our results regarding a positive association between FQ use and AA/AD corroborate prior reports, despite different study methods. 11-13, 15 Lee et al used a case-crossover design and found increased odds of exposure to FQs during the hazard period (60 days prior to AA/AD event) vs the referent period (one of three randomly selected 60-day periods between 120 and 300 days prior to AA/ AD event) (OR 2.15; 95%CI 1.14-6.46). 15 Two additional studies evaluated the risk of AA/AD with FQ exposure vs non-exposure periods, and the conclusions were the same. 11 Note: IRR, incidence rate ratio; SCCSA, self-controlled case series analysis; SMX-TMP, sulfamethoxazole-trimethoprim. a Adjusted for time-varying covariates of age, fiscal year, and respiratory, urinary, and skin and soft-tissue infections.
b Rows for fluoroquinolones vs "no antibiotics," "other antibiotics," and "multiple antibiotics" were removed so the sum of the events does not equal the total listed for each outcome (full results in Table S2). c The numbers in the rows that follow "fluoroquinolone risk period" are for the comparator antibiotics (eg amoxicillin, azithromycin). d N = 56 patients were removed due to only one observation period. was not part of the meta-analysis, did not find an association between levofloxacin and AMI after adjusting for a wide range of potential confounders. 10 For the outcome of VenA, results published in the literature comparing FQs with other antibiotics have also been mixed. 8,9,18,19 In the previously mentioned cohort study in Veterans

| CON CLUS ION
We found that FQs were associated with an increased risk of AA/ AD vs both amoxicillin and azithromycin and an increased risk of allcause mortality vs many antibiotics commonly used for outpatient infections. Although the differences in event rates are small, FQs should be reserved for serious infections where there are no suitable alternatives.

DATA S H A R I N G A N D DATA ACCE SS I B I LIT Y
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.

ACK N OWLED G M ENTS
There was no funding for this project. In kind support was provided