Lack of association between fluoroquinolone and aortic aneurysm or dissection

Abstract

Background and Aims

An increased risk of aortic aneurysm and aortic dissection (AA/AD) has been reported with fluoroquinolone (FQ) use. However, recent studies suggested confounding factors by indication. This study aimed to investigate the risk of AA/AD associated with FQ use.

Methods

This nationwide population-based study included adults aged ≥20 years who received a prescription of oral FQ or third-generation cephalosporins (3GC) during outpatient visits from 2005 to 2016. Data source was the National Health Insurance Service reimbursement database. The primary outcome was hospitalization or in-hospital death with a primary diagnosis of AA/AD. A self-controlled case series (SCCS) and Cox proportional hazards model were used. Self-controlled case series compared the incidence of the primary outcome in the risk period vs. the control periods.

Results

A total of 954 308 patients (777 109 with FQ and 177 199 with 3GC use) were included. The incidence rate ratios for AA/AD between the risk period and the pre-risk period were higher in the 3GC group [11.000; 95% confidence interval (CI) 1.420–85.200] compared to the FQ group (2.000; 95% CI 0.970–4.124). The overall incidence of AA/AD among the patients who received FQ and 3GC was 5.40 and 8.47 per 100 000 person-years. There was no significant difference in the risk between the two groups (adjusted hazard ratio 0.752; 95% CI 0.515–1.100) in the inverse probability of treatment-weighted Cox proportional hazards model. Subgroup and sensitivity analysis showed consistent results.

Conclusions

There was no significant difference in the risk of AA/AD in patients who were administered oral FQ compared to those administered 3GC. The study findings suggest that the use of FQ should not be deterred when clinically indicated.

Structured Graphical abstract

FQ, fluoroquinolone; 3GC, 3rd generation cephalosporin; HR, hazard ratio; CI, confidence interval; IPTW, inverse probability of treatment weighting.

Fluoroquinolones, Antibiotics, Aortic aneurysm, Aortic dissection

See the editorial comment for this article ‘Concerns about aortic aneurysm or dissection risk should not cease fluoroquinolone use when clinically indicated’, by J. Tamargo and S. Agewall, https://doi.org/10.1093/eurheartj/ehad660.

Introduction

Fluoroquinolones are a class of antibiotics that exert antibacterial activity by inhibiting bacterial DNA synthesis. Fluoroquinolone is one of the most commonly used antibiotic classes worldwide, owing to its unique advantages.¹ Broad antimicrobial spectrum, particularly including Pseudomonas aeruginosa, makes fluoroquinolone a useful agent for various community and hospital-acquired infections.^2–4 Furthermore, high bioavailability and excellent tissue penetration of fluoroquinolone allow oral treatment of difficult-to-eradicate infections which have traditionally been treated with a prolonged course of intravenous regimen.^5–7

Recent reports using a large healthcare usage database suggested that fluoroquinolone use might be associated with rare adverse events, particularly aortic aneurysm and aortic dissection (AA/AD).^8–10 Subsequently, the United States Food & Drug Administration (US FDA) issued a warning regarding the increased risk of AA/AD with fluoroquinolone use, which led to a reduction in the administration of fluoroquinolone.^11,12 However, there has been controversy regarding the biologic plausibility of vascular events occurring within a relatively short time period after fluoroquinolone use.¹³ Furthermore, recent studies reported that the indications for antibiotic use, rather than the antibiotic regimen itself, were more closely associated with the occurrence of vascular events.^14–16 Concerns include that clinicians may defer using fluoroquinolone owing to the small increased risk of AA/AD even when the use of fluoroquinolone provides a substantial benefit for the treatment of indicated infections.

Confirmation of the risk of vascular events associated with fluoroquinolone and estimation of the risk size would be of paramount importance in public health considering the wide use of this antibiotic class. Thus, we conducted a study to estimate the risk of AA/AD associated with fluoroquinolone, using large-scale nationwide healthcare usage data.

Methods

Data source

We used the healthcare reimbursement data from the National Health Insurance Service (NHIS) of the Republic of Korea. The NHIS is a universal single payer for all citizens and qualified long-term residents in Korea that reimburses all medical costs covered by the government-operated policy. The extracted data included age, sex, diagnoses coded according to the 10th edition of the International Classification of Diseases and Related Health Problems (ICD-10), and prescription of medications covered by the NHIS.

Study design

The study cohort comprised adults aged ≥20 years who received a prescription of the study drugs, oral fluoroquinolone, or third-generation cephalosporin (3GC), in outpatient visits from January 2005 to December 2016. The study period was set to conclude in 2016 as the new warning issued by the US FDA might have changed the prescription pattern of fluoroquinolone.¹⁷ We selected 3GC as a control because of its broad antimicrobial spectrum, clinical indications (including acute bacterial respiratory, gastrointestinal, and genitourinary infections), and favourable toxicity profile that matches fluoroquinolone (see Supplementary data online, Table S1 for comparison). The pharmaceutical codes used to identify the study drugs are listed in Supplementary data online, Table S2. The index date was the day on which the study drug was first prescribed. Only the first use of either of the study drugs for each patient was included; thus, all patients were represented only once in this study. Patients who met the following criteria were excluded: (i) previous diagnosis of primary outcome events before the study period, (ii) hospitalization during the previous 120 days from the index date, and (iii) concurrent use of both fluoroquinolone and 3GC. Any diagnosis within 3 years of the index date was included as comorbidity (see Supplementary data online, Table S3 for ICD-10 codes). Concomitant use of other drugs was identified by any prescription within 1 year of the index date.

The primary outcome was hospitalization or in-hospital death with a primary diagnosis of AA/AD (see Supplementary data online, Table S4 for the ICD-10 codes used to identify the primary outcome). We used two strategies to examine the association between study drug exposure and the primary outcome: (i) Cox proportional hazard model and (ii) a self-controlled case series (SCCS). In the Cox proportional hazard model, the risk for the primary outcome that occurred up to 1 year after the index date was directly compared between fluoroquinolone and 3GC. The patients who had diagnoses of AA/AD before the index date were excluded to mitigate the risk of previous diagnoses being carried over to the observation period. For the SCCS analysis, the incidence of the primary outcome was compared between the risk interval (7–187 days since the index date), the pre-risk period (6–12 months before the index date), and the post-risk period (6–12 months since the index date) within each group. The patients who had previous occurrences of primary outcome events within 1 year before the study period were excluded in SCCS.

Inverse probability of treatment weighting (IPTW) and propensity score (PS) matching were used for Cox proportional hazard model and SCCS, respectively, to further balance baseline characteristics. Inverse probability of treatment weighting was selected over PS matching for the Cox proportional hazard model as IPTW resulted in better balance (see Supplementary data online, Figure S1). Aortic aneurysm and aortic dissection may be found incidentally by advanced imaging (e.g. computed tomography or magnetic resonance imaging) performed to diagnose the indications for antibiotics. Also, AA/AD may cause a catastrophic event that leads to out-of-hospital death. Thus, we conducted sensitivity analyses with (i) surgical or vascular interventions for AA/AD, (ii) outpatient diagnosis of AA/AD, and (iii) out-of-hospital death as outcome events.

Statistical analysis

Propensity score was calculated using multiple logistic regression with fluoroquinolone use as a dependent variable and sex, age, comorbidities, concomitant drugs, and focus of infection as independent variables. Propensity score matching was conducted using the greedy nearest neighbour matching method in a 1:1 ratio. Inverse probability of treatment weighting was performed by weighting the patients with stabilized weights: [proportion of fluoroquinolone prescribed]/PS for the patients who used fluoroquinolone and [proportion 3GC prescribed]/(1 − PS) for those who used 3GC.¹⁸ Cox proportional hazards model was constructed using the PHREG procedure, and robust sandwich-type variance estimators were used to estimate corrected variance and confidence intervals (CI). Incidence rate ratios (IRR) for SCCS were calculated through Poisson regression using GENMOD procedure in SAS.

All tests were two-tailed, and statistical significance was set at P < .05. Analyses were performed using SAS Enterprise Guide 7.15 (SAS Institute, Cary, NC, USA). The SAS code used for analysis is provided as a Supplementary data online.

Results

Study population

A total of 954 308 patients were identified and formed the overall cohort for AA/AD (Figure 1). Fluoroquinolone was 4.39 times more commonly prescribed than 3GC. Patients in the fluoroquinolone group were slightly older (median, 42 years vs. 41 years; P < .001) and more frequently female (54.23% vs. 53.02%; P < .001), but the distribution of comorbidities was generally balanced with absolute differences of <1 percentage point (%p; Table 1 and Supplementary data online, Table S5). Similarly, the difference in concomitant drugs between the two groups was mostly <1%p, although some drugs (anxiolytic, oral glucocorticoid, non-steroidal anti-inflammatory drug, opiate, and proton pump inhibitor) showed a 1%–5%p difference between the two groups. However, there were substantial differences in the distribution of the infection foci. Patients with respiratory, skin/soft tissue, and bone/joint infections were more likely to receive 3GC while fluoroquinolone was more frequently prescribed for genitourinary and gastrointestinal infections.

Figure 1

Study flowchart. FQ, fluoroquinolone; 3GC, 3rd generation cephalosporin; IPTW, inverse probability of treatment weighting

Table 1

Baseline characteristics of overall and inverse probability of treatment-weighted cohorts

Characteristics	Unweighted cohort			Inverse probability of treatment-weighted cohort
	FQ (n = 777 109)	3GC (n = 177 199)	Absolute SMD	FQ (n = 897 062)	3GC (n = 907 927)	Absolute SMD
Sex
Male	355 697 (45.77)	83 244 (46.98)	0.024	408 934 (45.59)	397 548 (43.79)	0.036
Female	421 412 (54.23)	93 955 (53.02)		488 128 (54.41)	510 378 (56.21)
Age (median, IQR)	42 [30–55]	41 [30–54]	0.055	42 [30–55]	42 [30–55]	0.007
20–29	183 284 (23.59)	43 093 (24.32)		209 932 (23.40)	212 732 (23.43)
30–39	164 261 (21.14)	41 995 (23.70)		194 145 (21.64)	192 393 (21.19)
40–49	154 702 (19.91)	33 101 (18.68)		178 241 (19.87)	176 207 (19.41)
50–59	134 353 (17.29)	29 653 (16.73)		154 603 (17.23)	159 357 (17.55)
60–69	87 280 (11.23)	18 589 (10.49)		99 793 (11.12)	103 374 (11.39)
≥70	53 229 (6.85)	10 768 (6.08)		60 347 (6.73)	63 863 (7.03)
Comorbidities
Diabetes	90 490 (11.64)	20 034 (11.31)	0.011	104 445 (11.64)	108 713 (11.97)	0.010
Hypertension	138 314 (17.80)	30 135 (17.01)	0.021	159 498 (11.78)	166 331 (18.32)	0.014
Ischaemic heart disease	41 350 (5.32)	9236 (5.21)	0.005	48 472 (5.40)	49 915 (5.50)	0.004
Heart failure/cardiomyopathy	14 496 (1.87)	3399 (1.92)	0.004	17 015 (1.90)	17 659 (1.94)	0.004
Valvular heart disease	1087 (0.14)	313 (0.18)	0.009	1338 (0.15)	1360 (0.15)	<0.001
Cerebrovascular disease	33 932 (4.37)	7994 (4.51)	0.007	39 973 (4.46)	41 482 (4.57)	0.005
Arterial disease	55 216 (7.11)	11 746 (6.63)	0.019	63 145 (7.04)	65 589 (7.22)	0.007
Arrhythmia	20 707 (2.66)	4715 (2.66)	<0.001	24 319 (2.71)	24 761 (2.73)	0.001
Chronic lung disease	115 021 (14.80)	30 193 (17.04)	0.061	136 449 (15.21)	136 055 (14.99)	0.006
Cancer	16 382 (2.11)	4334 (2.45)	0.023	19 679 (2.19)	20 004 (2.20)	0.001
Chronic liver disease	77 263 (9.94)	16 203 (9.14)	0.027	89 683 (10.00)	91 953 (10.13)	0.004
Kidney disease	20 793 (2.68)	4477 (2.53)	0.009	24 080 (2.68)	25 523 (2.81)	0.008
Rheumatologic disease	34 950 (4.50)	7133 (4.03)	0.023	40 490 (4.51)	40 917 (4.51)	<0.001
Psychiatric disorder	170 528 (21.94)	37 019 (20.89)	0.026	196 819 (21.94)	202 495 (22.30)	0.009
Neurologic disorder	21 486 (2.76)	5791 (3.27)	0.029	25 425 (2.83)	26 932 (2.97)	0.008
Peripheral neuropathies and myopathies	46 240 (5.95)	9578 (5.41)	0.024	53 244 (5.94)	54 953 (6.05)	0.005
Concomitant drugs
Non-study antibiotics	45 056 (5.80)	9843 (5.55)	0.011	51 364 (5.73)	54 026 (5.95)	0.010
Angiotensin-converting enzyme inhibitor or angiotensin receptor blocker	64 093 (8.25)	15 290 (8.63)	0.014	73 757 (8.22)	77 149 (8.50)	0.010
Calcium channel blocker	68 017 (8.75)	13 961 (7.88)	0.032	78 935 (8.80)	82 929 (9.13)	0.012
Loop diuretic	7361 (0.95)	1812 (1.02)	0.008	8765 (0.98)	8800 (0.97)	0.001
Other diuretic	33 923 (4.37)	6305 (3.56)	0.041	39 059 (4.35)	40 796 (4.49)	0.007
Beta-blocker	44 498 (5.73)	9210 (5.20)	0.023	51 461 (5.74)	52 342 (5.76)	0.001
Digoxin	2279 (0.29)	502 (0.28)	0.002	2702 (0.30)	2883 (0.32)	0.003
Nitrates	6876 (0.88)	1667 (0.94)	0.006	8154 (0.91)	8424 (0.93)	0.002
Platelet inhibitor	47 259 (6.08)	11 286 (6.37)	0.012	55 100 (6.14)	57 911 (6.38)	0.010
Anticoagulant	1401 (0.18)	438 (0.25)	0.014	1678 (0.19)	2228 (0.25)	0.013
Lipid-lowering agent	51 969 (6.69)	12 576 (7.10)	0.016	59 151 (6.59)	62 139 (6.84)	0.010
Focus of infection
Lower respiratory infections	10 544 (1.36)	5698 (3.22)	0.125	14 729 (1.64)	15 203 (1.67)	0.002
Upper respiratory infections	206 309 (26.55)	64 758 (36.55)	0.216	254 343 (28.35)	254 356 (28.02)	0.007
Genitourinary tract infections	137 968 (17.75)	5440 (3.07)	0.495	134 013 (14.94)	147 958 (16.30)	0.046
Skin and skin structure infections; bone and joint infections	21 429 (2.76)	7568 (4.27)	0.082	27 459 (3.06)	28 032 (3.09)	0.001
Gastrointestinal infections	69 768 (8.98)	2663 (1.50)	0.340	66 382 (7.40)	69 360 (7.64)	0.011
Intra-abdominal infections; cardiovascular infections; septicaemia	6043 (0.78)	1674 (0.94)	0.018	7379 (0.82)	8927 (0.98)	0.017

FQ, fluoroquinolone; 3GC, 3rd generation cephalosporin; SMD, standardized mean difference; IQR, interquartile range.

The PS-matched cohorts included 317 972 patients with AA/AD. The patients in each group were well-matched in terms of sex and age; as a result, the differences in comorbidities, concomitant drugs, and the focus of infection were nominally larger in some variables. However, the difference did not exceed 5%p except for chronic lung disease, oral glucocorticoid use, and the focus of infections. In comparison, IPTW resulted in a superior balance between the two groups, with the absolute standardized mean differences of all variables being <.1 (see Supplementary data online, Table S6 and Figure S1).

Cox proportional hazard model

The overall incidence of AA/AD among patients who received fluoroquinolone and 3GC was 5.40 and 8.47 per 100 000 person-years, respectively. There was no statistically significant difference in the risk between the two groups [adjusted hazard ratio (aHR) 0.686; 95% CI 0.366–1.286; Table 2 and Supplementary data online, Table S7]. Adjusted analysis with IPTW also demonstrated comparable risk of AA/AD between the two groups (aHR, 0.752; 95% CI 0.515–1.100). Among the various subgroups, only patients aged <65 years had significant difference in risk; fluoroquinolone was associated with a lower risk of AA/AD in the weighted analysis (aHR 0.465; 95% CI 0.240–0.901; see Supplementary data online, Table S8 for the full model). Large variability in the incidence was observed by the focus of infection, ranging from 0 to 52.65 per 100 000 person-year. The incidence of AA/AD was highest in the patients with lower respiratory tract infection in both groups. The number of the patients who used 3GC for genitourinary, skin and soft tissue, and gastrointestinal infections were too small (<10 000) to detect the occurrence of AA/AD. However, no significant association with antibiotics was found in any indications.

Table 2

Risk of aortic aneurysm/aortic dissection by antibiotic use in the Cox proportional hazard model

Outcome (per 100 000 person-year)	Unweighted				Weighted
	FQ (n = 777 109)	3GC (n = 177 199)	Adjusted HR (95% CI)	P	FQ (n = 897 062)	3GC (n = 907 927)	Adjusted HR (95% CI)	P
Overall	5.40	8.47	.686 (0.366–1.286)	.240	5.13	7.01	0.752 (0.515–1.100)	.142
Subgroup
Male	8.72	14.42	0.655 (0.323–1.327)	.240	7.98	11.92	0.747 (0.483–1.154)	.188
Female	2.61	3.19	0.694 (0.180–2.680)	.596	2.74	3.18	0.727 (0.345–1.532)	.402
Age ≥65	33.26	42.45	0.919 (0.400–2.112)	.843	31.46	31.06	0.982 (0.604–1.597)	.941
Age <65	1.61	4.41	0.376 (0.137–1.030)	.057	1.60	3.66	0.465 (0.240–0.901)	.023
Indication for antibiotics
Lower respiratory	9.48	52.65	0.174 (0.020–1.556)	.118	9.17	58.54	0.152 (0.022–1.041)	.055
Upper respiratory	4.85	6.18	0.520 (0.201–1.346)	.178	4.28	7.84	0.703 (0.347–1.426)	.329
Genitourinary	1.45	0.00	N/A		1.55	0.00	N/A
Skin and soft tissue/bone	4.67	0.00	N/A		5.34	0.00	N/A
Gastrointestinal	4.30	0.00	N/A		4.66	0.00	N/A
Other	7.74	8.69	0.924 (0.360–2.373)	.870	7.04	8.48	0.856 (0.521–1.406)	.538

The numbers of patients in each subgroup are shown in Table 1.

CI, confidence interval; FQ, fluoroquinolone; 3GC, 3rd generation cephalosporin.

Self-controlled case series

The IRR for AA/AD between the risk period and the pre-risk period were 2.000 (95% CI 0.970–4.124) and 11.000 (95% CI 1.420–85.200) among the patients who received fluoroquinolone and 3GC, respectively (Table 3). While the IRR was 5.5 times higher in the 3GC group, the 95% CIs of the incidence rates overlapped. No significant increase in the incidence was observed between the risk and post-risk periods. The analyses of the PS-matched cohorts yielded results consistent with those of the overall cohorts; only the patients with 3GC use showed a significantly increased risk of AA/AD during the risk period compared to the pre-risk period (IRR 10.000; 95% CI 1.280–78.116).

Table 3

Incidence of aortic dissection/aortic aneurysm in self-controlled case series

Cohort	Antibiotic	Number of events			Incidence rate ratio (95% confidence interval)
		Pre-risk period	Risk period	Post-risk period	Risk period vs. pre-risk period	P	Risk period vs. post-risk period	P
Overall	FQ (n = 777 228)	11	22	21	2.000 (0.970–4.124)	.061	1.048 (0.576–1.905)	.879
	3GC (n = 177 226)	1	11	6	11.000 (1.420–85.200)	0.022	1.833 (0.678–4.957)	.234
PSM	FQ (n = 158 992)	1	3	3	3.000 (0.312–28.841)	.341	1.000 (0.202–4.955)	>.999
	3GC (n = 158 992)	1	10	5	10.000 (1.280–78.116)	.028	2.000 (0.684–5.852)	.206

FQ, fluoroquinolone; PS, propensity score; 3GC, 3rd generation cephalosporin; PSM, propensity score matched.

Sensitivity analysis

In the sensitivity analysis using the Cox proportional hazard model with the PS-matched cohort, no significant difference in the risk for AA/AD was observed (Table 4A). When alternative outcome events were examined, fluoroquinolone use was significantly associated with a lower risk of surgery or vascular intervention for AA/AD in both unweighted and weighted analyses (Table 4B). Outpatient diagnosis of AA/AD was also lower in the fluoroquinolone group (aHR 0.546; 95% CI 0.429–0.695). No significant difference in out-of-hospital death was observed by antibiotic use.

Table 4

Sensitivity analysis

(A) Cox proportional hazard model with the propensity score-matched cohort
	Overall				Propensity score-matched
	FQ (n = 777 109)	3GC (n = 177 199)	Adjusted HR (95% CI)	P	FQ (n = 158 986)	3GC (n = 158 986)	Adjusted HR (95% CI)	P
Outcome (per 100 000 person-year)	5.40	8.47	0.686 (0.366–1.286)	.638	1.03	7.34	1.324 (0.715–2.453)	.372

(B) Cox proportional hazard model with alternative outcome events
Outcome (per 100 000 person-year)	Unweighted				Weighted
	FQ (n = 777 109)	3GC (n = 177 199)	Adjusted HR (95% CI)	P	FQ	3GC	Adjusted HR (95% CI)	P
Surgery or radiologic intervention	2.57	6.77	0.383 (0.171–0.858)	.020	2.42	6.33	0.393 (0.238–0.650)	<.001
Outpatient diagnosis of AA/AD	12.10	17.49	0.655 (0.426–1.008)	.055	11.47	21.41	0.546 (0.429–0.695)	<.001
Out-of-hospital death	5.15	7.34	1.319 (0.681–2.554)	.411	5.09	4.44	1.172 (0.770–1.785)	.459

CI, confidence interval; FQ, fluoroquinolone; 3GC, 3rd generation cephalosporin; AA/AD, aortic aneurysm/aortic dissection.

Discussion

In this large study using nationwide healthcare reimbursement data, no difference was identified in the risk for AA/AD between individuals who were administered fluoroquinolone and those who were administered 3GC. The incidence of AA/AD was nominally higher in the 3GC group; however, the difference was not statistically significant (Structured Graphical Abstract).

A series of studies using Taiwanese national health insurance data suggested that fluoroquinolone use is associated with an increased risk of AA/AD.^8,19 Pasternak et al.⁹ also reported the increased risk associated with fluoroquinolone use compared to amoxicillin using a Swedish historical cohort. These studies led to an FDA warning about the ‘increased risk of ruptures or tears in the aorta blood vessel’ with fluoroquinolone use.¹¹ This warning possibly resulted in the decrease in prescribing fluoroquinolone in both inpatient and outpatient settings.^12,20 Various subsequent studies reaffirmed the initial findings that fluoroquinolone use is associated with an increased incidence of AA/AD.^10,21–24 Increased matrix metalloproteinase by fluoroquinolone and subsequent degradation of collagen was proposed as a potential mechanism behind this finding.^25,26

Despite the strong association from multiple studies, some argue that it is implausible that fluoroquinolone causes AA in such a short time frame as AA is a condition that progresses slowly.¹³ Also, concerns have been raised that fluoroquinolone may be avoided even when clinically indicated and the benefit greatly overweighs the small absolute risk of AA/AD. A recent meta-analysis estimated that 7246 patients should be treated with fluoroquinolone to cause one excess case of AA/AD.²⁷

Previous studies reporting the association between fluoroquinolone use and AA/AD share an important limitation of inadequate comparators. Some studies applied a self-controlled design or control that did not receive antibiotics.^8,19,24 A self-controlled design cannot distinguish the effect of infections treated with fluoroquinolone from that of fluoroquinolone itself, as they are essentially concurrent events. Controls with no antibiotic use have a fundamental difference in that they did not have infections. Infection has been suspected to be a risk factor for AA due to bacterial invasion of the arterial wall, haemodynamic instability, and systemic inflammation. Furthermore, advanced imaging used to diagnose the focus of infection may lead to incidental discovery of asymptomatic AA. The results from the SCCS analysis in our study confirmed this by showing that the use of both antibiotics was associated with an increased risk of AA/AD. Thus, the incidence of vascular events after fluoroquinolone use should be compared with the use of other antibiotics that share clinical indications and usage patterns with fluoroquinolone. Most studies that compared fluoroquinolone with other antibiotics selected amoxicillin, azithromycin, and trimethoprim-sulfamethoxazole as control drugs, as they share FDA-approved indications with fluoroquinolone.^9,10,21–23 However, these comparators have limited clinical application in the real world. Amoxicillin is generally used for mild bacterial respiratory infections with a low risk of resistance.^2,28 Azithromycin is now questioned as an empirical treatment for pneumonia due to an increase in resistance² and the use of trimethoprim-sulfamethoxazole for urinary tract infections require caution owing to its high resistance rate.⁴ Thus, antibiotics used as controls in most studies do not quite seem to be comparable to fluoroquinolone in terms of clinical indications and antibacterial spectrum. As the relevant clinical decision is almost always ‘which antibiotics to use,’ rather than ‘whether to use antibiotics,’ inadequate comparators undermine the applicability of the study findings.

Dong et al.¹⁴ examined the possibility of indication bias in a robust case–control study. They reported that fluoroquinolone was not associated with an increased risk of AA/AD compared to β-lactam/β-lactam inhibitors or extended-spectrum cephalosporins. Rather, the type of indicated infection was strongly associated with an increased risk of AA/AD, septicaemia and intra-abdominal infection having the highest risk. Similarly, subsequent studies have also shown that fluoroquinolone is not associated with AA/AD when the indication is limited to a single category of infection (urinary tract infection) or prophylaxis.^16,29 Our results are in line with the latter studies in that we used a comparator that is similar to fluoroquinolone in terms of clinical indication and antibacterial spectrum. The incidences of AA/AD in our study were also markedly different by the focus of infection, from 0 to 52.65 per 100 000 person-years, suggesting the impact of infection itself on the development of AA/AD. Subsequent studies on this subject should include infection focus in the analysis.

In our study, the incidence of primary outcome events was low, 5.97 per 100 000 person-years for AA/AD. Thus, even if there exists a small increase in the relative risk associated with the choice of antibiotic that could not be verified in this study, the absolute increase in risk would be exceedingly small, as the incidence of AA/AD is low. In comparison, the optimal treatment of infectious diseases for which antibiotics are indicated would affect survival, morbidity, and quality of life to a substantially greater degree. Fluoroquinolone has certain advantages compared to many classes of antibiotics: excellent bioavailability, high tissue penetration into certain body compartments (e.g. bone and joint, central nervous system), and a broad spectrum. These advantages support the use of fluoroquinolone for osteoarticular infections and infections involving prosthetic devices, which are notorious for their difficulty in treatment and poor outcome when treatment fails.^5,6 Thus, our results strongly suggest that the use of fluoroquinolone, when clinically indicated, should not be deterred for the concern of vascular adverse events.

Our study had several advantages. First, we used nationwide healthcare reimbursement data with a forward design. As the NHIS is a universal single payer in the Korean healthcare system, its database is comprehensive and reliable. All antibiotics are prescription drugs in Korea; therefore, our capture of exposure was considered exhaustive. Furthermore, our dataset and design enabled us to examine the incidence of the primary outcome measures, and thus, the absolute risk. Second, we applied several strategies to improve the robustness of our study. The SCCS is a useful design that minimizes the effect of interpersonal variability in baseline characteristics, as all patients serve as their own controls. The Cox proportional hazards model provided a direct comparison of the risk by antibiotic use. Furthermore, we used PS-based matching or weighting to mitigate residual bias. Lastly, a sensitivity analysis was conducted with multiple alternative outcome measures. The initial diagnosis of vascular incidents may have been carried over to subsequent hospital admissions. While we limited the capture of ICD-10 codes to primary diagnoses, the results from the sensitivity analysis reaffirmed those of the main analysis.

However, our study had several limitations. Our study was observational in nature; we utilized two different study designs with PS-based methods, but the possibility of residual bias and imbalance remains. The difference in some baseline characteristics could not be reduced below the threshold for statistical significance because of the large size of the dataset. We also excluded individuals diagnosed with AA/AD that preceded the study period. Although we found such exclusion necessary to minimize the risk of those diagnoses being carried over into the study period, this limits the generalizability of our findings. Further studies are warranted in patients with a history of vascular events.

In conclusion, there was no significant difference in the risk of AA/AD in patients who were administered oral fluoroquinolone compared to those administered 3GC. Third-generation cephalosporins was associated with a higher risk of AA/AD in the elderly. Our findings strongly suggest that the use of fluoroquinolone should not be deterred when clinically indicated.

Supplementary Data

Supplementary data are available at European Heart Journal online.

Declarations

Disclosure of Interest

All authors declare no conflict of interest for this contribution.

Data Availability

The data underlying this article were accessed from the National Health Insurance Service of the Republic of Korea, which cannot be shared publicly due to the Personal Information Protection Act (Act No. 11990).

Funding

Gil Medical Center, Gachon University (2019-11) and National Research Foundation of Korea (NRF-2021R1A5A2030333) to M.K. and J.J. The sponsor was not involved in the study design or in the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the paper.

Ethical Approval

The study protocol was approved by the Institutional Review Board at the Gachon University Gil Medical Center (approval number GCIRB2021-051). Consent was waived by the ethics committee because the data involved routinely collected medical data that was processed anonymously at all stages.

Pre-registered Clinical Trial Number

Not applicable.