Evaluating the Fluoroquinolone–Aortic Aneurysm and Dissection Signal: From Detection to Regulatory Action

A pharmacovigilance signal is an evidence state, not a permanent conclusion. The fluoroquinolone–aortic aneurysm and dissection history illustrates this particularly well because the regulatory position evolved through several stages: an initial epidemiological concern, an early conclusion that evidence was insufficient for product-information change, additional observational and non-clinical evidence, a 2018 PRAC recommendation to amend product information, and later studies that reopened questions about confounding by infection and comparator choice.

This article is deliberately process-focused. The related QPPV.com article [[fluoroquinolones-and-aortic-aneurysm-dissection-signal-evaluation]] examines the clinical and epidemiological evidence as a signal-evaluation case study. Here, the emphasis is on how the conclusion changed over time and how a pharmacovigilance professional should reconstruct that change without hindsight bias.

Purpose and Scope

The central question is not whether one can attach a single timeless label—"causal" or "not causal"—to the association. The more instructive questions are:

• What evidence existed when the concern first entered formal signal management?
• Why was that evidence initially considered insufficient for a label change?
• What new evidence altered the balance by 2018?
• What exactly did PRAC recommend, and what did that recommendation mean?
• How should later evidence that attenuates the association affect the current scientific interpretation?

The case therefore teaches a general rule: historical regulatory action and current causal interpretation must be kept distinct. A decision can have been proportionate to the evidence available at the time even if later evidence changes the estimated magnitude or certainty of the risk.

Define the Safety Question Before Reviewing the Evidence

Aortic aneurysm and aortic dissection are related but distinct vascular outcomes. An aneurysm is abnormal dilatation of the aorta; dissection involves separation of layers of the aortic wall after an intimal tear. Combining them can be useful when events are rare, but it can also obscure different associations for the component outcomes.

The exposure question also needs precision. The European signal concerned fluoroquinolones for systemic use and later included inhaled use in the 2018 product-information recommendation. A class-level regulatory conclusion should not be interpreted as evidence that every active substance has an identical quantitative risk.

A useful signal statement at the outset is therefore:

Recent exposure to systemic fluoroquinolones may be associated with an increased risk of aortic aneurysm and/or dissection, with uncertainty about causality, magnitude, susceptible populations and the relative contribution of the two component outcomes.

That wording preserves what was being investigated without assuming the answer.

The First Regulatory Stage: Concern Without Label Change

Observational reports published in 2015 raised concern that fluoroquinolone exposure might be associated with serious collagen-related adverse outcomes, including aortic aneurysm or dissection. The hypothesis was biologically plausible because fluoroquinolones were already associated with tendon injury and experimental work suggested potential effects on extracellular-matrix integrity. Plausibility, however, was only one part of the assessment.

PRAC discussed the signal in 2016 under EPITT 18651. In its September 2016 minutes, PRAC recorded that it had considered the studies by Daneman and colleagues and Lee and colleagues together with marketing-authorisation-holder responses. At that stage, PRAC concluded that the available evidence was insufficient to warrant an update of product information and agreed that routine pharmacovigilance should continue.

This stage is important because it demonstrates what "signal" means operationally. The issue was credible enough to evaluate formally, but the evidence had not yet crossed the threshold for a regulatory label change.

Why the early evidence was difficult to interpret

The early studies were observational. Patients receiving fluoroquinolones differed from unexposed patients in clinically important ways. They had infections, different comorbidity burdens and different patterns of healthcare contact. Aortic disease is also strongly associated with age, hypertension, atherosclerosis, smoking and connective-tissue disorders.

Consequently, an observed association could reflect some combination of:

• a true medicine-related effect;
• the infection for which the antibiotic was prescribed;
• characteristics influencing antibiotic choice;
• greater diagnostic imaging and detection;
• residual confounding after statistical adjustment.

The correct interpretation was therefore neither dismissal nor causal certainty. Continued surveillance was a proportionate response to an unresolved but potentially serious concern.

What Changed by 2018?

Two developments mattered. First, further epidemiological evidence appeared, including the Swedish nationwide cohort study by Pasternak and colleagues. This study used amoxicillin as an active comparator, an important methodological improvement over comparison with untreated populations because both groups required antibiotic treatment.

Second, additional non-clinical evidence supported biological plausibility. PRAC's September 2018 signal recommendation explicitly records consideration of epidemiological studies by Lee, Daneman and Pasternak, non-clinical work by LeMaire and colleagues, and MAH responses.

The evidence had therefore become broader and more convergent than it had been in 2016. That did not remove all uncertainty, but it changed the regulatory balance.

The 2018 PRAC Recommendation

On 6 September 2018, PRAC recommended that marketing authorisation holders of fluoroquinolones for systemic and inhaled use submit a variation within three months to amend product information. The proposed SmPC wording stated that epidemiological studies reported an increased risk of aortic aneurysm and dissection after fluoroquinolone exposure, particularly in older people, and advised careful benefit-risk assessment in patients with predisposing conditions or risk factors.

The wording is instructive. It did not say that every exposed patient was at risk or that a precise causal effect had been established. It described the epidemiological evidence, identified susceptible populations and translated the concern into targeted risk communication.

This distinction matters because a regulatory warning is not merely a scientific conclusion written in legal form. It is a risk-management intervention designed around the seriousness of the outcome, uncertainty in the evidence, patient susceptibility and the availability of alternative antibacterial treatments.

Reassessing a Signal After Regulatory Action

A common weakness in historical safety writing is to stop at the first label change. Pharmacovigilance does not stop there. Once a risk has been communicated, later evidence can strengthen, narrow, broaden or weaken the underlying causal model.

For the fluoroquinolone–aortic association, later observational work placed increasing emphasis on two methodological problems: confounding by infection and choice of comparator.

Infection as a competing explanation

An antibiotic is prescribed because a patient has an infection. If the infection itself is associated with vascular events, inflammation, hospitalisation or greater diagnostic investigation, then infection can create an apparent relationship between antibiotic exposure and the outcome.

This is not merely a statistical nuisance. It changes the causal question. The relevant counterfactual is usually not:

What would have happened if this patient had received no antibiotic?

but rather:

What would have happened if a clinically comparable patient with a similar infection had received another appropriate antibiotic?

That is why active-comparator designs are especially important in medicine-safety epidemiology.

Later Evidence and Attenuation of the Association

Dong and colleagues examined a large population using a design intended to address infection more directly. In that analysis, infections themselves were associated with aortic aneurysm or dissection. Among patients with relevant infections, fluoroquinolone exposure did not show a clear excess compared with selected alternative antibiotics after adjustment.

Gopalakrishnan and colleagues similarly showed that estimates could differ by clinical indication and comparator. Analyses addressing imaging and other sources of surveillance difference attenuated some associations.

These studies did not mathematically prove that the fluoroquinolone effect was zero. They showed that the estimate was sensitive to study design and control of plausible bias. That is a materially different scientific position from a stable association reproduced across increasingly rigorous designs.

What attenuation means

When a safety association weakens after better confounding control, three interpretations remain possible:

1. much of the original association was non-causal;
2. a smaller causal effect remains after removal of bias;
3. methodological changes introduce different biases and obscure a true effect.

The evaluator must therefore examine the complete design, not simply compare point estimates.

Aneurysm and Dissection Should Not Be Forced Into One Phenotype

Later evidence also highlighted the value of separating the component outcomes. Several reviews have reported a more consistent association for aneurysm than for dissection. This does not automatically invalidate the original combined signal. Combined outcomes are often used because rare events create statistical imprecision. However, once enough data exist, continuing to treat clinically distinct outcomes as one phenotype can conceal important differences.

A mature signal record should therefore be capable of changing from:

aortic aneurysm or dissection

toward a more granular conclusion if the evidence supports it.

Signal refinement is not evidence that the original process failed. It is evidence that the process is working.

Relationship to the Broader Fluoroquinolone Referral

The aortic signal should also be distinguished from the broader EU Article 31 referral concerning quinolone and fluoroquinolone medicines that concluded with restrictions because of disabling and potentially permanent adverse effects involving other organ systems. The fact that both processes concerned fluoroquinolones does not make them the same safety question.

This is an important reading discipline. A product-information change can arise from a signal procedure while a broader referral is simultaneously evaluating a different set of risks. When reconstructing history, the PV professional should map each conclusion to its own regulatory source and legal pathway.

How to Build a Defensible Historical Signal Assessment

A useful reconstruction contains four layers.

1. Contemporaneous evidence

Document what was known at each decision point. For this case, 2016 and 2018 should not be collapsed into one stage because PRAC reached different regulatory conclusions as the evidence changed.

2. Regulatory interpretation

Record exactly what PRAC concluded and what action it recommended. Avoid rewriting measured language into stronger causal claims.

3. Later scientific evidence

Identify studies that both support and challenge the original hypothesis. Later evidence belongs in the current interpretation but should not be projected backwards into earlier decision-making.

4. Current operational consequence

Check the current authorised product information and applicable restrictions. Historical articles explain how a risk was evaluated; they should not substitute for current prescribing information.

Practical Signal-Evaluation Framework

For a similar observational safety signal, an experienced PV team would normally examine the following connected questions:

No single item decides causality. The strength of the signal lies in how these elements converge.

Benefit-Risk and Proportionality

Regulatory action under uncertainty is often misunderstood as a choice between "proven" and "unproven." In practice, the decision is proportional.

A potential aortic event is serious and may be fatal. Some patients have substantially elevated baseline risk. Alternative antibiotics may be available. A warning and advice to consider alternatives in susceptible patients impose a different burden from withdrawal of an effective antibacterial class.

Therefore, the threshold of evidence needed for a targeted warning can reasonably be lower than the threshold that would be expected for market withdrawal. This is not a relaxation of scientific standards; it is the application of benefit-risk reasoning to the type and reversibility of the regulatory intervention.

Failure Modes in Historical Signal Reconstruction

Several errors can make an otherwise detailed signal evaluation misleading.

The first is hindsight substitution: using evidence published after a regulatory decision to imply that the earlier decision was obviously right or wrong. Historical adequacy should be judged using the evidence available at the time.

The second is warning-as-proof reasoning. A regulatory warning establishes that the evidence justified communication and risk management; it does not transform an observational association into an exact causal risk estimate.

The third is confirmation-only review. Once a warning exists, later negative or attenuating studies may be unconsciously treated as less credible. A functioning pharmacovigilance system should actively seek evidence capable of changing its prior conclusion.

The fourth is procedure conflation. The aortic signal procedure and the broader Article 31 fluoroquinolone referral addressed overlapping products but different regulatory questions. Their conclusions should not be merged without tracing the source.

The fifth is phenotype inertia. If later evidence differentiates aneurysm from dissection, the signal should be refined rather than preserved indefinitely as an undifferentiated combined endpoint.

Inspection and Governance Perspective

For a mature signal that has already generated regulatory action, inspection attention would reasonably focus on whether the company continued to manage the evidence rather than treating the first label change as closure. Illustrative questions include:

• Can the organisation reconstruct why the conclusion differed between 2016 and 2018?
• Are regulatory decisions linked to the evidence and procedure that produced them?
• Were later epidemiological studies that weakened or refined the association identified and assessed?
• Can the signal record distinguish aortic aneurysm from aortic dissection where the evidence permits?
• Are comparator choice, infection and surveillance bias considered when interpreting observational studies?
• Is current product information reconciled with the internal safety position?
• Can the QPPV explain whether the current risk model differs from the historical one and why?

These are illustrative inspection questions, not reported findings or additional legal requirements.

Key Takeaways

The fluoroquinolone–aortic history is valuable because the regulatory conclusion did not move in a straight line from uncertainty to permanent certainty.

In 2016, PRAC considered the available evidence insufficient to warrant a product-information update. By September 2018, additional epidemiological and non-clinical evidence led PRAC to recommend a variation adding warnings and risk-based precautions. Later studies then showed how strongly the apparent association could depend on infection, comparator selection and surveillance.

The durable lessons are:

• preserve the evidence available at each regulatory decision point;
• distinguish signal status from causal certainty;
• interpret product-information changes as risk-management decisions, not numerical causal estimates;
• use active comparators and careful confounding control for observational drug-safety questions;
• refine combined outcomes when the evidence for their components diverges;
• continue scientific surveillance after regulatory action;
• judge historical decisions and current conclusions separately.

A high-quality pharmacovigilance system is therefore not one whose conclusions never change. It is one in which changes can be explained by new evidence, documented reasoning and proportionate regulatory action.

References

1. European Medicines Agency. Minutes of the PRAC meeting, 26–29 September 2016. EPITT 18651: fluoroquinolones — aortic aneurysm and dissection. EMA/PRAC/123588/2017.
2. 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. EMA/PRAC/595691/2018. Adopted 6 September 2018.
3. Pasternak B, Inghammar M, Svanström H. Fluoroquinolone use and risk of aortic aneurysm and dissection: nationwide cohort study. BMJ. 2018;360:k678.
4. Lee CC, Lee MTG, Chen YS, et al. Risk of aortic dissection and aortic aneurysm in patients taking oral fluoroquinolone. JAMA Internal Medicine. 2015;175(11):1839–1847.
5. Daneman N, Lu H, Redelmeier DA. Fluoroquinolones and collagen associated severe adverse events: a longitudinal cohort study. BMJ Open. 2015;5:e010077.
6. Dong YH, Chang CH, Wang JL, et al. Association of infections and use of fluoroquinolones with the risk of aortic aneurysm or aortic dissection. JAMA Internal Medicine. 2020.
7. Gopalakrishnan C, Bykov K, Fischer MA, Connolly JG, Gagne JJ, Fralick M. Association of fluoroquinolones with the risk of aortic aneurysm or aortic dissection. JAMA Internal Medicine. 2020.
8. European Medicines Agency. Quinolone- and fluoroquinolone-containing medicinal products — Article 31 referral. PRAC recommendation 2018; European Commission final decision 2019.

Regulatory Note

PRAC signal recommendations are pharmacovigilance regulatory recommendations. The legal implementation of a product-information change depends on the applicable marketing-authorisation and variation procedures. The separate quinolone/fluoroquinolone Article 31 referral had its own formal regulatory pathway and should not be treated as identical to EPITT 18651.

The epidemiological studies cited here are scientific evidence rather than legal requirements. Current prescribing and risk-management decisions should use current authorised product information and applicable guidance. The analytical framework and illustrative inspection questions are recommended explanatory practice.