- Design
- Unsupervised clustering of spontaneous adverse event reports (FAERS)
- Population
- 40,127 fluoroquinolone reports; comparators 39,948 macrolide and 16,310 amoxicillin reports
- Primary outcome
- Multi-system adverse event clusters
- Effect
- Cluster 10.1% of reports, 43.0% disabling vs 11.7%; signature in 14.9% vs 5.3% and 3.9%
This study applied unsupervised clustering to 40,127 adult reports in the FDA Adverse Event Reporting System, 2004 to 2025, in which a fluoroquinolone was the primary suspect drug, with macrolide and amoxicillin reports as comparators.
The algorithms consistently separated a multi-system cluster of 4,072 reports (10.1%) enriched for neuropsychiatric, musculoskeletal and sensory events — memory impairment, muscle atrophy and dry eye among the strongest terms. It held when seriousness and outcome were excluded from the clustering. Post hoc, 43.0% of cluster reports were disabling against 11.7% of the rest. The same signature matched 14.9% of fluoroquinolone reports, 5.3% of macrolide and 3.9% of amoxicillin reports.
The pattern resembles the disabling, potentially persistent multi-system reactions already in fluoroquinolone labelling and regulatory warnings. Spontaneous reports cannot give incidence or prove causation, and publicity can drive reporting. The value is methodological, and as a reminder that the restrictions on fluoroquinolone use for minor infections have a basis.
- Avoid fluoroquinolones for uncomplicated infections where alternatives exist, in line with regulatory advice.
- Warn patients to stop and report tendon pain, neuropathic symptoms or mood change on a fluoroquinolone.
- Consider fluoroquinolone exposure in a patient with unexplained multi-system symptoms after a course.
- Report suspected fluoroquinolone reactions through the national pharmacovigilance programme.
Why it matters
It shows the multi-system reaction pattern emerging from data alone, not only from clinical case reports.
Don't overread it
Spontaneous reporting data are hypothesis-generating; this does not measure how often the reaction occurs.
The statistics, in plain English
A lift of 8.6 for memory impairment means that term was 8.6 times as common in the cluster as in all fluoroquinolone reports. These are proportions of reports, not of patients treated.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for clinical pharmacology, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free