Findings from a recent study published in American Journal of Ophthalmology mined the FDA's spontaneous reporting system to map the ocular safety signals tied to three approved fibroblast growth factor receptor inhibitors (FGFRis).
Several of the signals reach beyond what current labeling captures, and the authors say they support proactive eye monitoring as these drugs move into wider use.
Give me some background first.
Some context: FGFRis—erdafitinib, pemigatinib, and futibatinib—treat FGFR-altered cancers such as urothelial carcinoma and cholangiocarcinoma.
As their use expands, so does attention to class-specific toxicity. Ocular adverse events (AEs) are among the most frequently reported, especially dry eye and a range of corneal and retinal abnormalities.
- And clinical trials have even documented visual disturbances in up to 25% of treated patients, including blurred vision, dry eye, retinal detachment, and keratitis.
But here's the issue: Investigators noted that trial populations are small and tightly screened, so they likely undercount the real-world spectrum.
- Additionally, the FDA labels list only a narrow set of ocular AEs, which leaves room for post-marketing data to surface events that pre-approval studies missed.
Now, talk about the study.
A team ran a retrospective pharmacovigilance analysis of the FDA Adverse Event Reporting System (FAERS) through OpenVigil 2.1, pulling reports where erdafitinib, pemigatinib, or futibatinib was the primary suspect drug.
They then calculated reporting odds ratios (RORs) for each drug-ocular AE pair against all other drugs in the database, covering April 19, 2019 to June 30, 2025.
- Keep in mind: An ROR reflects how often an event is reported for a drug relative to everything else in FAERS. It is not a measure of incidence, relative risk, or causation.
Who was included in the study?
The pull captured 1,582 unique FGFRi-linked adverse events. Of those, 200 (12.6%) were ocular.
Demographics: 562 reports came from male patients, 414 from female, and 606 didn't specify sex. Mean age was 63.7 years. The U.S. filed most reports (73.9%), followed by France (8.2%) and Canada (2.5%).
As for the drugs: Erdafitinib accounted for the bulk of ocular events at 150 (75%), versus 44 (22%) for pemigatinib and 6 (3%) for futibatinib.
So what were the findings?
Erdafitinib carried the broadest and strongest signals.The highest were corneal thinning (ROR 321; 95% CI 102 to 1009), ocular toxicity (ROR 225; 95% CI 117 to 439), and xerophthalmia (ROR 138; 95% CI 44 to 432). By raw count, dry eye led (n=31), then eye disorder (n=24), retinal detachment (n=15), and visual impairment (n=14).
Pemigatinib skewed toward retinal fluid problems. Its top signals were serous retinal detachment (RD; ROR 180; 95% CI 67 to 484), subretinal fluid (ROR 165; 95% CI 68 to 398), and retinal pigment epithelium detachment (RPED; ROR 110; 95% CI 35 to 344).
- Dry eye was, again, the most reported by count (n=20).
Meanwhile, futibatinib generated a single ocular AE signal: dry eye (n=6), at an ROR of 36.12 (95% CI 15.76 to 82.74).
Tell me more.
Many of the erdafitinib signals don't appear on its FDA label, including xerophthalmia, ulcerative keratitis (ROR 46), corneal thinning, chorioretinopathy (ROR 121), and maculopathy.
For pemigatinib: The events beyond the label included subretinal fluid, serous RD, RD, and ocular hyperemia—which, notably,was the lone event that fell short of statistical significance after Bayesian shrinkage.
On the labels themselves, erdafitinib lists dry eye in 28% of patients, pemigatinib lists dry eye in 31%, and futibatinib lists dry eye in 15%. The FAERS dry-eye reporting lined up with all three.
Were there any mismatches?
Just one: Futibatinib's label flags RPED in 9% of patients, but the analysis found no RPED signal for it.
The authors attributed this to FAERS capturing mainly symptomatic events, while small RPEDs often stay asymptomatic and resolve on their own.
Any limitations to consider?
The usual pharmacovigilance caveats apply. RORs are sensitive to small cell counts, so rare events can produce inflated, unstable estimates—a real concern in this case, given that several signals rested on three to five reports.
Other considerations:
- FAERS is subject to underreporting, duplicate entries, and missing clinical detail—and it cannot establish incidence, severity, or causality.
- With polypharmacy common in oncology, isolating the responsible drug is difficult—raising the risk of confounding.
- Erdafitinib was approved first (2019), giving it a longer surveillance window than pemigatinib (2020) or futibatinib (2022). That partly explains its larger share of reports, and a lower report count shouldn't be read as lower toxicity.
The authors framed the results as hypothesis-generating rather than confirmatory.
Expert opinion?
The authors argued the signals justify proactive ophthalmic monitoring for patients on FGFRis.
On mechanism, they noted these are pan-FGFR agents, and FGFR1-4 sit throughout the retina and concentrate in the retinal pigment epithelium, which may predispose those structures to toxicity.
They also pointed to erdafitinib's longer elimination half-life (roughly 59 hours, versus 15.4 for pemigatinib and 2.9 for futibatinib) as a plausible reason it dominates the corneal signals.
Anything else?
A meibomian gland angle fits prior work. Pemigatinib has been shown to drive meibomian gland atrophy and faster tear film breakup, which could explain the dry eye and compensatory tearing seen across all three agents.
Take home.
For clinicians managing patients on FGFR inhibitors, early counseling that dry eye and increased lacrimation are common, and that more serious retinal events can occur but usually resolve once the drug is stopped.
As such: Eyecare providers should match erdafitinib for corneal and retinal toxicity and pemigatinib for retinal fluid and serous detachments.
Baseline and ongoing eye monitoring is a reasonable default while these drugs see broader clinical use.