Antibiotic-associated encephalopathy: A retrospective pharmacovigilance study based on 21 years of data from the US FDA Adverse Event Reporting System

Abstract

Antibiotic-associated encephalopathy (AAE) is a serious adverse drug reaction that can cause significant neurological complications. This study aimed to identify the antibiotics associated with AAE by mining data from the US Food and Drug Administration Adverse Event Reporting System (FAERS), thereby providing insights into antibiotic-related neurotoxicity risks. Disproportionality analysis was performed using the reporting odds ratio method based on FAERS data from Q1 2004 to Q4 2024. A total of 87 preferred terms potentially related to AAE were identified. Antibiotics classified under Anatomical Therapeutic Chemical codes J01 to J07 and other antimicrobial agents were included. All identified antibiotics were grouped into 3 categories according to their clinical neurotoxic profiles. In total, 13,698 patients with AAE were identified over the 21-year period. The 5 most frequently reported antibiotics were metronidazole (1464 cases), levofloxacin (1447), ciprofloxacin (1290), ertapenem (825), and imipenem/cilastatin (788). The incidence of AAE increased with age. Among the 121 antibiotics analyzed, 52 showed positive reporting odds ratio signals: 28 were classified as type 1, 6 as type 2, and 18 as type 3. Most AAE cases occurred within 10 days of antibiotic exposure, except for several agents, including meropenem/vaborbactam and ticarcillin/clavulanic acid (type 1), and isoniazid, pyrazinamide, rifampicin, and paromomycin (type 3), which exhibited delayed onset. This study represents the first large-scale pharmacovigilance analysis of AAE using FAERS data and identifies 52 antibiotics with significant neurotoxicity signals. Most antibiotics trigger early-onset AAE, and older patients demonstrate greater susceptibility. These findings expand our understanding of AAE and inform safer antibiotic use.

1. Introduction

Antibiotics are among the most commonly prescribed medications worldwide and play an essential role in the treatment of bacterial infections. Although generally well tolerated, antibiotics can cause adverse events (AEs), including allergic reactions, rashes, and gastrointestinal disturbances.^[1] However, antibiotic-associated encephalopathy (AAE) is often under-recognized because of its nonspecific clinical manifestations. AAE has been described as a spectrum of neurotoxic effects induced by antibiotics and is categorized into 3 types based on clinical presentation: type 1 is characterized by seizures and myoclonus, type 2 by psychiatric symptoms, and type 3 by cerebellar ataxia.^[2] Although AAE typically resolves following timely adjustment of antimicrobial therapy, its underlying pathogenesis remains unclear, making it difficult to determine whether neurological symptoms are caused by antibiotic exposure or confounding factors such as fever or sepsis.^[3] Therefore, the clinical features of AAE must be characterized to facilitate timely identification and appropriate management.

The epidemiological burden of AAE has been increasingly recognized in recent years. It has been reported that the incidence of severe central nervous system adverse reactions induced by antibiotics is <1%.^[4,5] In a retrospective study, antibiotic use was implicated in 12 of 117 patients with status epilepticus.^[6] In a randomized trial involving more than 2500 hospitalized patients, delirium or coma was observed in 21% of the cefepime group and 17% of the piperacillin/tazobactam group.^[7] In another retrospective study of 100 intensive care unit patients, a 15% incidence of neurotoxicity was reported following cefepime administration, with increased susceptibility among those with chronic kidney disease.^[8] In addition, an incidence of 4.4% was documented in a study of 2104 patients with end-stage kidney disease, and poor outcomes were associated with AAE.^[9] These findings highlight the clinical significance of AAE, particularly in vulnerable populations. Despite growing awareness, most existing studies on AAE have been limited to small case series or have focused on narrow-spectrum antibiotics and isolated symptoms. Large-scale, comprehensive investigations systematically evaluating a wide range of antibiotics and their neurotoxic profiles remain lacking.

The US Food and Drug Administration Adverse Event Reporting System (FAERS) is a spontaneous reporting database regarded as one of the largest pharmacovigilance resources. It is designed to support post-marketing surveillance of drug safety by collecting suspected AE reports submitted by healthcare professionals, consumers, and manufacturers. The FAERS serves as a key tool for detecting rare or unexpected adverse drug reactions that may not be identified in premarketing clinical trials due to limited sample sizes and controlled conditions. Real-world data derived from such reporting systems have been widely utilized in pharmacovigilance and risk assessment.^[10]

To date, no comprehensive pharmacovigilance analysis of AAE has been conducted using the FAERS database. This study aimed to perform a disproportionality analysis of AAE cases reported in the FAERS and identify the antimicrobial agents associated with AAE. The findings are expected to enhance the understanding of AAE risk and support evidence-based, safer antibiotic use in clinical practice.

2. Materials and methods

2.1. Data source

Data were extracted from the FAERS database (https://fis.fda.gov/extensions/FPD-QDE-FAERS/FPD-QDE-FAERS.html; accessed April 2, 2025), which covers Q1 2004 to Q4 2024.

2.2. Target AEs and target medications

Given the heterogeneity of clinical manifestations associated with AAE, the Standardized MedDRA Queries category “noninfectious encephalopathy/delirium” was deemed insufficient to capture all relevant cases. To ensure comprehensive case identification and maintain the rigor of pharmacovigilance analysis, a systematic literature review was conducted, and additional AAE-related diagnostic terms were incorporated. AEs clearly attributed to non-drug-related pathological conditions (e.g., infections, brain trauma, and epilepsy) were excluded. Consequently, potentially relevant preferred terms (PTs) were identified and are listed in Table S1, Supplemental Digital Content, https://links.lww.com/MD/R726.

After AEs were queried based on 87 PTs, antibiotics were screened. According to the Anatomical Therapeutic Chemical (ATC) classification system, both anti-infectives for systemic use (codes J01–J07) and other antimicrobial agents, including those categorized under A07 (anti-infectives for systemic use in the gastrointestinal tract), were included.

2.3. Data extraction and processing

Data extraction and processing were conducted in strict accordance with the US Food and Drug Administration (FDA) guidelines. The fields PRIMARYID, CASEID, and FDA_DT were retrieved from the DEMO table and used to organize the reports. When multiple reports shared the same CASEID, only the record with the most recent FDA_DT was retained for analysis. In cases where both CASEID and FDA_DT were identical, the report with the highest PRIMARYID score was preserved.

For data submitted after Q1 2019, deduplication was performed based on the deleted files provided with each quarterly dataset. Additional deletions were performed using CASEID listings from deletion report logs. Only reports in which the antibiotic was identified as the “primary suspect drug” were included in the analysis. All data were processed, analyzed, and visualized using SAS version 9.4 software (SAS Institute Inc., Cary).

2.4. AAE signal detection and statistical analysis

The disproportionality analysis method, known as the reporting odds ratio (ROR), was employed to detect AAE signals.^[11] The ROR was calculated using the formula $\mathrm{R}\mathrm{O}\mathrm{R}=\frac{a/c}{b/d}$, and the lower limit of the corresponding 95% confidence interval (95% CI) was determined as $95\mathrm{\%}\mathrm{C}\mathrm{I}=e^{In\left(ROR\right)\pm 1.96\sqrt{(\frac{1}{a}+\frac{1}{b}+\frac{1}{c}+\frac{1}{d})}}$. A risk signal was present if ROR > 1, the lower limit of the 95% CI > 1, and ROR ≥ 3. In these calculations, a refers to the number of AAE reports for a given antibiotic, b is the number of reports involving other AEs associated with that antibiotic, c is the number of reports involving 1 of the 87 PTs but associated with other drugs, and d is the number of reports involving other drugs and other AEs. Descriptive statistics for the count data are expressed as case numbers and proportions. All statistical analyses and data visualization were performed using SAS version 9.4.

2.5. Classification of AAE drugs

The clinical phenotypes of AAE are categorized into 3 distinct types. Accordingly, antibiotics associated with AAE were primarily classified into 3 corresponding groups to enhance the understanding of the pathophysiological mechanisms underlying different forms of encephalopathy.^[12,13] Antibiotics that could not be assigned to types 1 or 2 were categorized as type 3, although these agents may not share common clinical features or mechanisms of action. The classification of antibiotics into these 3 types is shown in Table 1.

Table 1.

Classification of medications for antibiotic‑associated encephalopathy.

Type	Characteristic	Mechanism	Medication
Type 1	Myoclonus, seizures, abnormal EEG, and normal MRI	Disruption of inhibitory synaptic transmission leading to excitotoxicity	β-lactam antibiotics, including penicillin and cephalosporins
Type 2	Psychosis and abnormal EEG	Perturbations of the D2 dopamine and NMDA glutamate receptors	Sulfonamides, fluoroquinolones, and macrolides
Type 3	Cerebellar dysfunction with abnormal MRI	Free radical formation and altered thiamin metabolism	Metronidazole
	Psychosis (common), seizures (rare), and EEG (frequently abnormal but nonspecifically) (2)	–	Isoniazid
	Others	–	Cycloserine, dapsone, ethambutol, pyrazinamide, rifampicin

EEG = electroencephalogram, MRI = magnetic resonance imaging.

3. Results

3.1. Basic information on patient characteristics of AAE

Between Q1 2004 and Q4 2024, 13,698 patients with AAE were identified in the FAERS database. Figure 1 illustrates the detailed study process. The basic patient characteristics are presented in Table 2. The reporting proportions were similar between females (45.80%) and males (45.17%), with 9.03% of cases lacking specified sex information. The median patient age was 57 years (interquartile range, 39–71 years).

Figure 1.

Flowchart of study process for identification of AAE reports. AAE = antibiotic-associated encephalopathy, AE = adverse event, DEMO = demographic information, DRUG = drug information, PS = primary suspect, PT = preferred terms, REAC = adverse events.

Table 2.

Basic information on patient characteristics of antibiotic-associated encephalopathy.

Characteristics	Patients with AAE (n = 13,698)
Gender	n (%)
Female	6274 (45.80)
Male	6187 (45.17)
Unknown	1237 (9.03)
Age	n
Median (IQR)	57.00 (39.00, 71.00)
Reporter’s occupation	n (%)
Physician	3906 (28.52)
Pharmacist	3785 (27.63)
Other health professional	3164 (23.10)
Consumer	1954 (14.26)
Lawyer	40 (0.29)
Unknown	849 (6.20)
Reporter country	n (%)
United States	4366 (31.87)
France	1475 (10.77)
United Kingdom	1296 (9.46)
China	1019 (7.44)
Japan	892 (6.51)
Not specified	742 (5.42)
Other country	3908 (28.53)
Seriousness	n (%)
Serious	13,526 (98.74)
Non-serious	172 (1.26)
Time to event	n/d
Patients (missing)	5733 (7965)
Mean (SD)	20.28 (145.76)
Median (Q1, Q3)	3.00 (1.00, 8.00)

AAE = antibiotic-associated encephalopathy, IQR = interquartile range, SD = standard deviation.

As shown in Figure 2, the incidence of AAE increased with advancing age. Most reports were submitted by healthcare professionals, and the United States accounted for the highest number of AAE reports. Among the reported cases, 98.74% were classified as serious cases. The median time to AAE onset was 3.00 days. A greater number of AAE cases were reported in the latter half of the study period than in the earlier half (Fig. 3). The clinical outcomes associated with AAE are shown in Figure 4.

Figure 2.

Age division of antibiotic-associated encephalopathy patients.

Figure 3.

Report quantity of antibiotic-associated encephalopathy by year.

Figure 4.

Outcome of antibiotic-associated encephalopathy.

3.2. The top 50 antibiotics in terms of the number of AAE reports

A total of 121 antibiotics were identified in association with AAE, and the top 50 antibiotics ranked by the number of AAE reports are shown in Figure 5. Among them, 32 antibiotics demonstrated positive ROR signals.

Figure 5.

Top 50 antibiotics in terms of the number of antibiotic-associated encephalopathy reports. CI = confidence interval, ROR = reporting odds ratio.

Based on the second-level ATC classification, 40 of the top 50 antibiotics were categorized as J01 drugs (antibacterials for systemic use). Six drugs – voriconazole, amphotericin B, posaconazole, itraconazole, caspofungin, and micafungin – were classified as J02 (antimycotics for systemic use). The remaining 4, rifaximin, isoniazid, terbinafine, and the combination of bismuth/metronidazole/tetracycline, were assigned to other ATC categories.

All top 10 antibiotics demonstrated a positive ROR signal. Theseincluded metronidazole (1464 cases; ROR: 5.02; 95% CI: 4.76–5.29), levofloxacin (1447 cases; ROR: 1.86; 95% CI: 1.77–1.96), ciprofloxacin (1290 cases; ROR: 1.25; 95% CI: 1.19–1.32), ertapenem (825 cases; ROR: 21.16; 95% CI: 19.69–22.75), imipenem/cilastatin (788 cases; ROR: 14.19; 95% CI: 13.20–15.26), moxifloxacin (588 cases; ROR: 1.60; 95% CI: 1.47–1.73), rifaximin (566 cases; ROR: 2.90; 95% CI: 2.67–3.15), cefepime (527 cases; ROR: 18.11; 95% CI: 16.56–19.81), clarithromycin (491 cases; ROR: 1.43; 95% CI: 1.31–1.56), and linezolid (490 cases; ROR: 2.41; 95% CI: 2.20–2.63).

3.3. Antibiotics with positive ROR signals and medication types associated with AAE

Antibiotics associated with AAE were categorized into 3 types based on their clinical classification. Beta-lactamase inhibitors were assigned to type 1 owing to their beta-lactam structure and their frequent co-administration with beta-lactam antibiotics. Antibiotics that did not meet the criteria for either type 1 or type 2 were classified as type 3, although they may not share common neurotoxic mechanisms or clinical features. A complete list of antibiotics with positive ROR signals and their corresponding classifications is presented in Table 3.

Table 3.

Antibiotics with positive ROR signals and medication types associated with antibiotic-associated encephalopathy.

Drug	Report	ROR (95% CI)	AAE types	MTE
Imipenem	14	48.16 (26.66–86.99)	1	3.00
Ertapenem	825	21.16 (19.69–22.75)	1	5.00
Cefepime	527	18.11 (16.56–19.81)	1	4.00
Cilastatin/imipenem	788	14.19 (13.2–15.26)	1	2.00
Doripenem	59	11.17 (8.58–14.53)	1	5.00
Cilastatin/imipenem/relebactam	6	7.80 (3.44–17.66)	1	4.50
Meropenem/sodium carbonate anhydrous	6	7.74 (3.42–17.53)	1	–
Avibactam	7	7.58 (3.56–16.15)	1	6.00
Ceftazidime	147	6.44 (5.47–7.60)	1	4.00
Avibactam/ceftazidime	73	6.13 (4.85–7.74)	1	4.00
Tazobactam	4	6.05 (2.23–16.39)	1	1.00
Isoniazid	268	5.92 (5.24–6.69)	3	102.00
Paromomycin	4	5.85 (2.16–15.83)	3	44.00
Metronidazole	1464	5.02 (4.76–5.29)	3	5.00
Meropenem	293	4.19 (3.73–4.70)	1	3.00
Cefoxitin	10	4.19 (2.24–7.84)	1	–
Meropenem/vaborbactam	3	4.18 (1.33–13.12)	1	18.00
Anidulafungin	24	4.09 (2.73–6.13)	3	0.00
Colistin	27	3.91 (2.67–5.72)	3	9.00
Cefditoren	37	3.83 (2.77–5.31)	1	6.00
Cefiderocol	15	3.55 (2.13–5.91)	1	6.00
Sarecycline	5	3.48 (1.44–8.43)	3	–
Cefoperazone	9	3.16 (1.63–6.10)	1	9.50
Clavulanic acid/ticarcillin	9	2.96 (1.53–5.71)	1	19.00
Cefotaxime	59	2.91 (2.25–3.76)	1	3.00
Rifaximin	566	2.90 (2.67–3.15)	3	87.00
Pyrazinamide	19	2.87 (1.82–4.51)	3	66.00
Gatifloxacin	76	2.77 (2.21–3.47)	2	1.00
Ceftolozane/tazobactam	25	2.73 (1.84–4.05)	1	8.00
Isoniazid/pyrazinamide/rifampicin	15	2.67 (1.60–4.44)	3	18.50
Bismuth/metronidazole/tetracycline	54	2.48 (1.90–3.25)	3	7.00
Cefoperazone/sulbactam	19	2.43 (1.55–3.83)	1	3.00
Cefazolin	88	2.42 (1.96–2.99)	1	0.50
Linezolid	490	2.41 (2.20–2.63)	3	4.00
Ceftriaxone	392	2.24 (2.03–2.48)	1	1.00
Voriconazole	444	2.18 (1.99–2.40)	3	3.00
Ofloxacin	81	2.16 (1.73–2.68)	2	5.00
Piperacillin/tazobactam	318	2.11 (1.89–2.36)	3	3.00
Ceftaroline fosamil	16	2.03 (1.24–3.32)	1	5.50
Levofloxacin	1447	1.86 (1.77–1.96)	2	3.00
Cefaclor	11	1.83 (1.01–3.31)	1	1.00
Caspofungin	54	1.62 (1.24–2.11)	3	3.00
Posaconazole	86	1.60 (1.29–1.98)	3	4.00
Moxifloxacin	588	1.60 (1.47–1.73)	2	1.00
Daptomycin	189	1.58 (1.37–1.82)	3	7.00
Ampicillin/sulbactam	30	1.50 (1.05–2.15)	1	3.00
Clarithromycin	491	1.43 (1.31–1.56)	2	2.00
Tigecycline	48	1.43 (1.07–1.90)	3	2.00
Amphotericin B	125	1.39 (1.16–1.65)	3	3.00
Benzylpenicillin	44	1.36 (1.01–1.83)	1	2.00
Ciprofloxacin	1290	1.25 (1.19–1.32)	2	3.00
Vancomycin	366	1.16 (1.05–1.29)	3	5.00

AAE = antibiotic-associated encephalopathy, CI = confidence interval, MTE = median time-to-event, ROR = reporting odds ratio.

A total of 52 antibiotics were significantly associated with AAE, as evidenced by positive ROR signals. Notably, all the top 10 antibiotics in terms of signal intensity were classified as type 1. Among the 52 identified antibiotics, 28 were categorized as type 1, 6 as type 2, and 18 as type 3.

Several classes of antimicrobial drugs were represented in the group of 18 type 3 agents. These included 2 metronidazole preparations (metronidazole and bismuth/metronidazole/tetracycline), 5 antifungal agents (anidulafungin, voriconazole, caspofungin, posaconazole, and amphotericin B), and 3 tetracycline class drugs (sarecycline, bismuth/metronidazole/tetracycline, and tigecycline). In addition, 3 antitubercular antibiotics were identified (isoniazid, pyrazinamide, and isoniazid/pyrazinamide/rifampicin) along with 6 other antimicrobial agents (paromomycin, rifaximin, colistin, linezolid, daptomycin, and vancomycin).

3.4. ROR analysis of all medications with AAE at the PT level

A total of 87 PTs related to AAE were identified and retrieved for analysis. However, no medications were reported in the FAERS database for 5 of these PTs. The results of the ROR analysis for 121 antibiotics across the remaining 82 PTs are shown in Figure 6.

Figure 6.

ROR analysis of all antibiotics with antibiotic‑associated encephalopathy at the PT level. The asterisk (*) indicates that the corresponding PT simultaneously meets the ROR threshold. PT = preferred terms, ROR = reporting odds ratio.

Each antibiotic was associated with 1 or more PTs. Most PTs had corresponding reports available. Several antibiotics have been implicated in several PTs, such as status epilepticus, seizures, delirium, generalized tonic-clonic seizure, and encephalopathy, indicating common neurological presentations associated with antimicrobial use. In contrast, certain PTs, such as opsoclonus myoclonus, vascular encephalopathy, and neonatal epileptic seizures, were associated with relatively few antibiotics.

In addition, 28 PTs, including epileptic psychosis, subacute myelo-opticoneuropathy, and alcoholic seizures, were not associated with any antibiotics in the FAERS database.

3.5. Identifying antibiotics with new potential AAE risk

Six antibiotics classified as type 1 were newly identified to be associated with AAE. These included cefditoren, cefiderocol (previously noted only in 1 FAERS-based study),^[14] ceftolozane/tazobactam, ceftaroline fosamil, cefaclor, and benzylpenicillin. For type 3 antibiotics, apart from previously reported agents such as isoniazid, metronidazole, pyrazinamide, rifampicin, and certain antifungal drugs including voriconazole^[15–17] and amphotericin B,^[18] all other antibiotics were newly identified in this analysis as having a potential risk of AAE occurrence.

3.6. Time-to-event analysis of AAE for antibiotics

The median time-to-event (MTE) for AAE onset among all antibiotics with positive ROR signals was 3.00 days (Fig. 7). The detailed MTE values for the individual antibiotics are presented in Table 3.

Figure 7.

Median time-to-event of antibiotic-associated encephalopathy. IQR = interquartile range.

Most antibiotics classified as type 1 and type 2 exhibited short onset times, typically within several days of treatment initiation. Exceptions included meropenem/vaborbactam and ticarcillin/clavulanic acid, which had delayed onset times of 18.00 and 19.00 days, respectively.

For the type 3 antibiotics, onset times varied considerably. Metronidazole and bismuth/metronidazole/tetracycline exhibited MTEs of 5.00 and 7.00 days, respectively. However, substantially longer onset times were observed for several antitubercular agents, including isoniazid (102.00 days), pyrazinamide (66.00 days), and the combination isoniazid/pyrazinamide/rifampicin (18.50 days). Other type 3 agents with prolonged MTEs included rifaximin (87.00 days) and paromomycin (44.00 days). In contrast, the remaining type 3 antibiotics, primarily antifungal agents, were associated with AAE onset within a few days, similar to the pattern observed for type 1 and type 2 medications.

4. Discussion

In this study, 21 years of data from FAERS were analyzed to conduct the most comprehensive pharmacovigilance evaluation of AAE to date. Three principal objectives were achieved: first, 87 AAE-related PTs were identified; second, 13,698 patients were systematically evaluated using disproportionality analysis, resulting in the identification of 52 antibiotics with statistically significant risk signals; third, the post-marketing characteristics of AAE were described. These findings not only support prior evidence but also reveal novel associations, providing important guidance for the identification and risk management of AAE during clinical anti-infective therapy.

Compared with a recent multicenter hospital-based cohort study by Kim et al,^[13] which evaluated the incidence and risk factors of AAE and demonstrated that combination antibiotic therapy and reduced renal function were significantly associated with a higher likelihood of AAE, our FAERS-based study cannot estimate the true incidence but captures a broader spectrum of implicated antimicrobial agents. This discrepancy reflects inherent methodological differences: prospective cohorts provide incidence estimates and adjusted risk assessments, whereas pharmacovigilance databases are better suited for large-scale signal detection and rare event identification. Therefore, our findings complement rather than contradict existing epidemiological evidence.

The epidemiological and clinical characteristics of AAE were examined. A comparable sex distribution was observed (female: 45.8%; male: 45.2%), with a median age of 57 years, consistent with previous studies reporting no sex-based predisposition to AAE.^[12] However, a progressive increase in AAE incidence was noted with advancing age, consistent with the findings of Mattappalil et al,^[5] who attributed increased vulnerability in older patients to cumulative drug exposure, polypharmacy, and age-related changes in blood–brain barrier integrity.^[19] Most reports (79.3%) were submitted by healthcare professionals, including physicians and pharmacists, which enhanced the reliability of the data. Importantly, 98.7% of AAE cases were classified as serious, with outcomes including prolonged hospitalization, disability, or life-threatening events, highlighting the critical need for timely recognition. In addition, a steady increase in annual case reports was observed, likely reflecting improved pharmacovigilance systems and growing awareness of drug-induced neurotoxicity.

The antibiotics most frequently reported in association with AAE were metronidazole (1464 cases), levofloxacin (1447), ciprofloxacin (1290), ertapenem (825), and imipenem/cilastatin (788), accounting for 37.6% of all AAE cases. These findings are consistent with previous studies identifying β-lactams and fluoroquinolones as high-risk neurotoxic agents.^[12] Our results corroborate the predominance of type 1 and type 2 agents while expanding the spectrum by identifying additional type 3 agents through disproportionality analysis. This suggests that the clinical boundaries of AAE may be broader than previously defined. Metronidazole was the most frequently reported antibiotic. Several mechanisms have been proposed to explain its neurotoxicity, including the inhibition of GABAergic (gamma-aminobutyric acid) neurotransmission, interference with neuronal RNA and protein synthesis, generation of free radicals, and mitochondrial dysfunction. Moreover, neuroimaging changes characteristic of metronidazole neurotoxicity have been described, although the precise epileptogenic mechanisms remain unclear.^[20–23] Reported incidence rates vary across studies; for instance, levofloxacin-associated AAE has been estimated to occur in 1% to 2% of exposed individuals.^[24] A systematic review of 292 publications revealed 391 AAE cases, of which 7.4% involved metronidazole, and 6.6% involved ciprofloxacin use.^[12] Among carbapenems, imipenem has been associated with seizure rates as high as 3% to 33%, whereas the rate for ertapenem is substantially lower at 0.18%.^[25] Discrepancies between literature-based case counts and our pharmacovigilance data may reflect differences in the study populations, publication bias, and inherent limitations of small-scale reports. In contrast, large pharmacovigilance databases capture broader trends and can detect rare signals with greater sensitivity. Therefore, integrating real-world evidence with published literature is recommended for comprehensive risk assessment and clinical decision-making.

Among the 52 antibiotics identified with significant ROR signals, the top 10 agents, all classified as type 1, showed the strongest associations with AAE. These include multiple carbapenems and third- or fourth-generation cephalosporins, commonly used in critically ill patients. AAE associated with these drugs typically presents as myoclonus or seizures within a few days of treatment initiation. Previous studies have reported a high prevalence of renal dysfunction in patients with cephalosporin-associated encephalopathy, including cefepime (72%), cefpirome (70%), and ceftazidime (92%). Renal impairment may enhance neurotoxicity through multiple mechanisms, such as reduced clearance leading to elevated plasma concentrations, hypoalbuminemia increasing the free drug fraction, and altered protein metabolism compromising blood–brain barrier function.^[12] This mechanistic explanation is supported by the hospital-based study by Kim et al,^[13] which identified renal dysfunction as an independent predictor, reinforcing the clinical relevance of dose adjustment strategies. Renal function-guided dosing and close neurological monitoring are therefore essential to minimize neurotoxicity risk. Early recognition of AAE is crucial to reducing complications, shortening hospitalization, and improving patient outcomes. The neurotoxicity of imipenem, ertapenem, cefepime, doripenem, and imipenem/cilastatin has been well documented.^[8,26–31] The β-lactam ring structure is believed to lower the seizure threshold through GABA receptor antagonism and enhanced excitatory neurotransmission. Structural features such as basic amino side chains at the C2 position in imipenem may further potentiate epileptogenic activity.^[20]

In addition to the traditionally recognized high-risk type 1 and type 2 antimicrobial agents, several type 3 antibiotics were identified, exhibiting positive AAE signals. These include antitubercular agents such as isoniazid, whose neurotoxicity is primarily attributed to the disruption of pyridoxal metabolism, resulting in impaired neurotransmitter synthesis.^[32] Antifungal agents, such as voriconazole, posaconazole, and amphotericin B, have also been implicated, with proposed mechanisms involving increased blood–brain barrier permeability.^[16,33–35] In particular, voriconazole plasma concentrations have been shown to strongly correlate with neurological AEs.^[36,37] Pharmacokinetic studies by Pascual et al^[36] and Dolton et al^[37] demonstrated a clear exposure-toxicity relationship, supporting therapeutic drug monitoring to mitigate neurological AEs. Furthermore, Kato et al ^[17] suggested dopaminergic pathway involvement in voriconazole-associated hallucinations, highlighting the mechanistic diversity across antimicrobial classes.^[17] Other identified antimicrobial classes included daptomycin, rifaximin, and colistin. Colistin-induced neurotoxicity has been linked to increased neuronal membrane permeability, inhibition of acetylcholine release, oxidative stress, and mitochondrial dysfunction.^[32,38] Notably, positive AAE signals were also detected for agents such as caspofungin and the tetracycline derivative tigecycline, which have not traditionally been regarded as neurotoxic agents. These findings suggest that drugs not conventionally viewed as neurotoxic may still pose risks for neurological AEs, particularly in susceptible populations, including patients with hepatic or renal dysfunction, prolonged antimicrobial exposure, and underlying central nervous system disorders. Accordingly, heightened clinical vigilance and individualized risk assessments are recommended when prescribing these agents.

Bhattacharyya et al^[12] reported that encephalopathy typically develops within a few days after the administration of type 1 and type 2 antibiotics, whereas encephalopathy induced by metronidazole or isoniazid may occur several weeks or months post-administration. In this study, type 1 and type 2 antibiotics constituted the majority of agents associated with AAE, and their onset times were consistent with prior findings, except for meropenem/vaborbactam (18.00 days) and ticarcillin/clavulanic acid (19.00 days), which exhibited a delayed onset. The MTE for AAE was 3.00 days, which aligns with previously described onset windows.^[12] Most AAE cases attributed to type 1 and type 2 antibiotics occurred within 7 days of treatment initiation, indicating that neurotoxic effects often manifest early and require close monitoring from the outset of treatment.

However, a distinct subset of antibiotics demonstrated a significantly delayed the onset of AAE. Isoniazid, rifaximin, and pyrazinamide exhibited markedly prolonged MTEs of 102, 87, and 66 days, respectively. Isoniazid has the longest latency. Previous studies have suggested that isoniazid-induced neurotoxicity may result from the progressive impairment of presynaptic GABA synthesis.^[12] Co-administration of vitamin B6 has been shown to mitigate these effects,^[32,39] although the precise mechanisms remain unclear. These observations highlight the importance of long-term neurological surveillance when prescribing antibiotics with delayed neurotoxic potential and emphasize the need for individualized risk-mitigation strategies.

Despite these insights, several limitations of this study should be acknowledged. The FAERS is a spontaneous reporting system that is subject to reporting bias, missing data, and variability in case details.^[40,41] Duplicate records and data quality variability may also affect reliability, although efforts were made to remove duplicates from the data.^[11,42] Important clinical variables, such as comorbidities, concomitant medications, and treatment duration, were not consistently available, potentially confounding the interpretation.^[43] In addition, detailed clinical data, such as electroencephalogram (EEG) findings, neuroimaging results, and laboratory values, were unavailable. EEG patterns are particularly relevant for characterizing the neurophysiological correlates of AAE.^[44] The absence of such data limits our ability to correlate specific antibiotics with their corresponding EEG manifestations.

Nevertheless, FAERS remains one of the most widely utilized tools for adverse drug reaction surveillance. Compared with smaller retrospective studies, their breadth and diversity allow the detection of rare and emerging drug-related risks.^[6,9,19] Although ROR analysis adjusts for background event rates, temporal and geographic variability in reporting practices may influence signal detection. Furthermore, the inclusion of 87 PTs to maximize sensitivity may have introduced heterogeneity in AAE definitions.

Future investigations should integrate electronic medical records, prospective clinical studies, and mechanistic experiments to validate these findings. Prospective studies incorporating comprehensive neurological assessments, including standardized EEG monitoring, neuroimaging, and cognitive evaluations, are needed to elucidate the electrophysiological and clinical correlates of AAE across antibiotic classes. Particular attention should be paid to high-risk populations, including older patients, individuals in intensive care settings, those receiving combination antimicrobial regimens, and patients with renal dysfunction.

5. Conclusion

This study presents the first large-scale pharmacovigilance analysis of AAE using 21 years’ worth of FAERS data. A total of 52 antibiotics were significantly associated with AAE, including both established high-risk agents (e.g., carbapenems, cephalosporins, and fluoroquinolones) and under-recognized neurotoxic drugs (e.g., caspofungin and tigecycline). Marked differences in the AAE onset times were observed across the drug types. Type 1 and type 2 agents typically trigger early-onset AAE, whereas drugs such as isoniazid exhibit delayed-onset neurotoxicity. Older patients were identified as a high-risk population requiring tailored monitoring and dose adjustment. These findings expand the known spectrum of AAE and provide practical insights into the safe use of antibiotics. Despite the limitations inherent in the FAERS, this study provides a foundation for future prospective studies, mechanistic validation, and risk prediction in vulnerable populations.

Acknowledgments

The authors acknowledge the US Food and Drug Administration (FDA) for granting free access to the FAERS database.

Author contributions

Conceptualization: Haiyan Wang, Jiali Zhang.

Data curation: Haiyan Wang.

Formal analysis: Huifang Jiang, Jiali Zhang.

Investigation: Huifang Jiang.

Methodology: Houci Yang, Jian Tang.

Software: Houci Yang, Jian Tang.

Supplementary Material

medi-105-e48395-s001.pdf ^{(237.4KB, pdf)}