Historical Signal Evaluation: Fluoroquinolones and Aortic Aneurysm or Dissection

The association between systemic fluoroquinolones and aortic aneurysm or dissection is a useful example of regulatory action under scientific uncertainty. The initial concern was driven mainly by observational epidemiology supported by biological plausibility rather than by a randomised trial or a distinctive cluster of spontaneous reports. The outcome was serious, the estimated relative effects in some studies were substantial, and potentially susceptible patients could be identified. PRAC therefore recommended product-information changes in 2018.

Subsequent research did not make the evidence uniformly stronger. Some meta-analyses continued to find an association, particularly for aortic aneurysm, while better-controlled active-comparator studies found substantial attenuation or no association after adjustment. That later evidence does not make the 2018 decision irrational. It demonstrates why a signal conclusion must remain revisable and why regulatory proportionality is assessed using the evidence available at the time.

Defining the Outcome Correctly

An aortic aneurysm is a pathological dilation of the aorta associated with weakening of the vessel wall. Aortic dissection occurs when a tear permits blood to enter and separate layers of the aortic wall. The two conditions share vascular risk factors and can be clinically related, but they are not the same outcome.

This distinction became increasingly important as evidence accumulated. Early studies and regulatory assessments often used a combined endpoint of aneurysm or dissection. Later analyses suggested that the association could be stronger and more consistent for aneurysm than for dissection. A signal evaluator should therefore preserve both levels of analysis:

• combined aortic aneurysm/dissection, when this matches the source study or regulatory signal;
• aneurysm and dissection separately, when the data permit refinement.

Combining outcomes can improve statistical power for rare events, but it can also mask clinically meaningful heterogeneity.

Why the Hypothesis Was Biologically Plausible

Fluoroquinolones were already associated with tendon disorders, including tendinopathy and tendon rupture. This raised the possibility that effects on extracellular matrix or connective-tissue homeostasis might extend beyond tendon. Experimental research subsequently described changes in collagen-related pathways and aortic wall integrity that provided additional plausibility.

Mechanistic coherence strengthens a signal when it points in the same direction as clinical evidence. It does not establish that an epidemiological association is causal. Aortic disease develops in patients with major baseline vascular risk, and infection itself can create physiological stress and increased healthcare contact. A plausible laboratory mechanism cannot remove those competing explanations.

The Epidemiological Signal

Several observational studies reported increased aortic aneurysm or dissection risk after recent fluoroquinolone exposure. These studies were influential because randomised antibiotic trials are generally too small and too short to evaluate uncommon vascular events. Large healthcare databases can include millions of treatment episodes and therefore provide statistical power that clinical trials lack.

The price of that scale is confounding. Fluoroquinolones are prescribed for infections, and people receiving them may differ systematically from those receiving no antibiotic or another antibiotic. Important differences can include age, infection type and severity, cardiovascular disease, hospitalisation, smoking, hypertension, renal disease, corticosteroid exposure and the likelihood of diagnostic imaging.

The core causal question is therefore not simply whether aortic events occur more often after fluoroquinolone exposure. It is whether they occur more often than they would after an appropriate alternative treatment in patients with a comparable indication and baseline risk.

Why comparator choice matters

Comparison with non-use can produce an apparently clean contrast, but it may compare infected patients receiving an antibiotic with people who are not currently ill enough to require one. An active comparator—another antibiotic used for a similar clinical indication—can reduce some confounding by indication and healthcare utilisation.

It does not eliminate confounding. Antibiotics are not interchangeable, and treatment choice may reflect allergy, infection site, resistance, renal function, severity or prior treatment. Nevertheless, the direction and magnitude of an association across several plausible comparators provide important information about robustness.

This signal therefore teaches a general principle of pharmacoepidemiology:

the quality of the counterfactual can matter more than the size of the database.

The Evidence Considered by PRAC in 2018

PRAC adopted a signal recommendation on 6 September 2018 for fluoroquinolones for systemic and inhaled use and aortic aneurysm/dissection (EPITT 18651). The Committee considered epidemiological studies, including work by Lee and colleagues, Daneman and colleagues and Pasternak and colleagues, non-clinical evidence including work by LeMaire and colleagues, and responses from marketing authorisation holders.

PRAC recommended that the marketing authorisation holders submit variations to amend product information. The proposed warning stated that epidemiological studies reported an increased risk of aortic aneurysm and dissection after fluoroquinolone exposure, particularly in older people. It advised careful benefit-risk assessment and consideration of other therapeutic options in patients with predisposing conditions or risk factors, and advised urgent medical attention for sudden abdominal, chest or back pain.

The wording is notable for what it did not do. It did not state that every exposed patient would develop aortic disease, that every individual fluoroquinolone had an identical quantified risk, or that observational association had become absolute causal certainty. It translated a serious and plausible concern into targeted clinical risk communication.

How Later Evidence Refined the Signal

A good historical signal evaluation has two clocks. The first asks what evidence was available when the regulator acted. The second asks how the association should be understood now. Mixing the two creates hindsight bias.

Later systematic reviews continued to report positive associations, but they also exposed important uncertainty. A 2022 systematic review and meta-analysis found an increased risk of de novo aortic aneurysm or dissection within 30 days of fluoroquinolone exposure (RR 1.42, 95% CI 1.11–1.81) and within 60 days (RR 1.44, 95% CI 1.26–1.64). The authors rated the certainty very low and low, respectively, and explicitly noted the possible effect of unmeasured confounding.

Another 2022 meta-analysis separated the two component outcomes. Across four observational studies eligible for pooling, fluoroquinolone use was associated with aortic aneurysm compared with other antibiotics (HR 1.84, 95% CI 1.10–2.48), whereas the pooled estimate for aortic dissection was not statistically significant (HR 1.09, 95% CI 0.96–1.25).

This does not mean that the original combined signal was “wrong.” It means the phenotype became more differentiated as more evidence accumulated. A mature signal process should allow precisely this kind of refinement.

Evidence that challenges the association

A large 2023 study using UK primary-care databases illustrates why confounding remains central. In crude cohort analyses, fluoroquinolone use was associated with hospitalisation for aortic aneurysm or dissection compared with cephalosporins. After adjustment for potential confounders, the pooled hazard ratio was 1.03 (95% CI 0.91–1.17). In a case-crossover analysis, an association was seen versus non-use but not versus several comparator antibiotics.

This is important evidence against a simplistic causal interpretation. It suggests that infection, treatment selection or related patient characteristics may explain some associations seen in less comparable groups. It does not, by itself, prove that there is no increased risk in any susceptible subgroup or with any exposure pattern.

The correct scientific response is therefore neither to ignore the study because it is reassuring nor to use it to erase the regulatory signal. It should reduce confidence in causal estimates that are not robust to comparator and confounding control.

Relative Risk, Absolute Risk and Baseline Susceptibility

Aortic aneurysm and dissection are uncommon in the general population but become substantially more likely with age and recognised vascular or connective-tissue disorders. PRAC's 2018 recommendation therefore focused particularly on patients with higher baseline risk, including people with pre-existing aneurysm or dissection, family history, Marfan syndrome, vascular Ehlers-Danlos syndrome, certain vasculitides, hypertension and known atherosclerosis.

This reflects a basic benefit-risk principle. If a medicine increased relative risk by the same proportion in two patients, the absolute increase would be larger in the patient with the higher baseline risk. Therefore, even an imprecise relative effect can be more consequential in a susceptible population.

For signal communication, relative estimates should not be presented without context. A two-fold association from an observational study can sound dramatic, but the actual number of additional events depends on the baseline rate and the validity of the causal estimate. Conversely, a low absolute frequency does not make aortic rupture or dissection clinically trivial.

Duration, Timing and Causal Interpretation

Several studies evaluated short windows after antibiotic exposure. A relatively close temporal relationship can support plausibility if a mechanism could operate over that time. It can also increase vulnerability to confounding by acute infection, because the exposure window overlaps the illness that prompted antibiotic treatment.

Reported duration-response relationships require similar caution. Longer fluoroquinolone courses may represent greater exposure, but they may also identify more severe or complicated infections and patients with different comorbidity. An apparent duration gradient is therefore supportive only after alternative explanations have been considered.

The evaluator should distinguish:

• pharmacological dose-response;
• treatment-duration association;
• cumulative exposure;
• acute indication severity;
• surveillance and imaging intensity.

These variables can move together in routine-care data.

Regulatory Action Under Uncertainty

The fluoroquinolone case demonstrates why the threshold for risk minimisation is not identical to the threshold for definitive causal proof. Regulators consider the seriousness of the outcome, the credibility and consistency of the evidence, biological plausibility, availability of alternative treatments, baseline susceptibility and the burden created by the proposed risk-minimisation measure.

A targeted warning can be proportionate when the potential harm is catastrophic and high-risk patients can be identified, even if the exact causal effect remains uncertain. Such an action should also remain open to re-evaluation if better evidence later changes the scientific assessment.

The key distinction is:

regulatory precaution based on an evidence-supported concern ≠ scientific declaration that all uncertainty has ended.

This distinction protects against two opposite errors: waiting for impossible certainty before communicating a serious plausible risk, and treating a precautionary warning as permanently conclusive evidence.

Practical Signal-Evaluation Framework

When an observationally driven serious-risk signal emerges, the pharmacovigilance team should build the assessment around several questions:

This framework is more useful than simply ranking studies by whether their confidence interval crosses one.

Failure Modes and Inspection Perspective

The first major failure is combined-endpoint inertia: continuing to describe aneurysm and dissection as though later evidence had not differentiated them. Signal terminology should evolve when the phenotype becomes clearer.

The second is comparator blindness. An association versus non-use may substantially attenuate when the comparison is made with another antibiotic. The evaluator should examine whether the apparent effect is robust to clinically credible counterfactuals rather than select the comparison producing the largest estimate.

The third is historical hindsight. Later studies that weaken an association should inform the current conclusion, but they should not be projected backwards as evidence that regulators possessed in 2018. The defensibility of a historical regulatory action is judged against the information and uncertainty available at the decision point.

The fourth is warning-as-proof reasoning. Product-information changes show that the concern was considered sufficient for regulatory risk communication; they do not convert an observational estimate into an immutable causal constant.

Illustrative inspection questions include:

• Can the organisation reconstruct the evidence and rationale underlying its class-level conclusion?
• Are aneurysm and dissection analysed separately when data permit?
• Does the assessment address indication and comparator choice rather than quoting adjusted effect estimates without design context?
• Are important later studies, including evidence that weakens the association, incorporated into ongoing evaluation?
• Are patients with high baseline aortic risk considered in benefit-risk reasoning without presenting recommended clinical practice as a new legal requirement?
• Is current product information reconciled with internal signal documentation and implemented through controlled change processes?

These are illustrative questions, not reported inspection findings.

Key Takeaways

The fluoroquinolone–aortic signal is best understood as a case study in action under uncertainty. In 2018, PRAC considered multiple epidemiological studies, non-clinical evidence and MAH responses and recommended product-information changes for systemic and inhaled fluoroquinolones. The seriousness of the outcome and the possibility of identifying higher-risk patients supported a targeted regulatory response.

Later evidence has made the causal picture less uniform. Some systematic reviews continue to report increased short-term associations. Other analyses suggest a stronger signal for aneurysm than dissection, and well-controlled active-comparator work has shown marked attenuation after confounding adjustment.

The durable lessons are:

• define combined outcomes carefully and refine them when component evidence diverges;
• scrutinise comparator choice and confounding in observational safety studies;
• distinguish biological plausibility from causal proof;
• interpret relative risk in the context of baseline susceptibility and absolute risk;
• judge historical actions using contemporaneous evidence while allowing the current scientific conclusion to evolve;
• recognise that proportionate risk minimisation can be justified before causal uncertainty is completely resolved.

The quality of a signal system is therefore demonstrated not by never changing its conclusion, but by changing it in a traceable and scientifically justified way when better evidence becomes available.

References

1. European Medicines Agency. PRAC recommendations on signals adopted at the 3–6 September 2018 PRAC meeting. Fluoroquinolones for systemic and inhaled use — aortic aneurysm and dissection, EPITT 18651. Date of adoption: 6 September 2018.
2. Pasternak B, Inghammar M, Svanström H. Fluoroquinolone use and risk of aortic aneurysm and dissection: nationwide cohort study. BMJ. 2018;360:k678.
3. Chen C, Patterson B, Simpson R, et al. Do fluoroquinolones increase aortic aneurysm or dissection incidence and mortality? A systematic review and meta-analysis. Frontiers in Cardiovascular Medicine. 2022;9:949538. doi:10.3389/fcvm.2022.949538. PMID: 36017083.
4. Fluoroquinolones and the risk of aortic aneurysm or aortic dissection: an updated systematic review and meta-analysis including 53,651,283 patients. 2022. PMID: 36468763.
5. Association Between Fluoroquinolone Use and Hospitalization With Aortic Aneurysm or Aortic Dissection. 2023. PMID: 37585175.

Regulatory Note

The 2018 PRAC signal recommendation was a regulatory pharmacovigilance recommendation for product-information variation. The precise legal implementation depends on the applicable marketing-authorisation procedures and subsequent authorised product information.

Observational studies and meta-analyses cited here are scientific evidence, not legal requirements. Current prescribing or clinical decisions should follow the current authorised product information and applicable clinical guidance for the specific medicine and jurisdiction. The analytical framework and illustrative inspection questions are recommended explanatory practice.