# Facial Skincare Adverse Event Atlas and Safety Signals From openFDA Cosmetic Reports: A Disproportionality Analysis

## TL;DR

This study maps adverse events from facial skincare products using FDA data and identifies safety signals, such as eye-area products causing ocular symptoms.

## Contribution

The paper introduces a facial skincare adverse event atlas and detects category-level safety signals using openFDA cosmetic reports.

## Key findings

- Eye-area products showed the strongest association with ocular symptoms (ROR 61.67).
- Leave-on products like masks and retinoids showed consistent signals for burn and swelling.
- Case-check analyses confirmed low product concentration, supporting category-level interpretations.

## Abstract

To construct a facial‐skincare adverse event atlas and detect category‐level safety signals using the US FDA openFDA Cosmetic Adverse Events database.

We analyzed publicly available openFDA cosmetic adverse event reports (downloaded 2025‐12‐15; meta.last_updated 2025‐12‐10). Suspect products were classified into 12 facial‐skincare subcategories using keyword rules, excluding hair, makeup, nail, fragrance, oral, deodorant, and bath/body cleansing products. MedDRA preferred terms were grouped into nine clinically interpretable reaction clusters. Reporting odds ratios (RORs) with 95% confidence intervals were computed versus the full database (primary) and within‐cohort comparators (sensitivity), with false discovery rate (FDR) control. We performed “case‐check” analyses for five prioritized signals to assess product concentration and temporal patterns.

The broad facial cohort comprised 2927 reports and the strict facial sensitivity cohort 1573 reports. The strongest full‐database signals were eye‐area products with ocular symptoms (ROR 61.67; 95% CI 46.13–82.43), masks with burn‐related events (ROR 27.95; 20.39–38.31), retinoid/anti‐aging with swelling/angioedema (ROR 21.94; 17.20–28.00), serum/essence with ocular symptoms (ROR 17.82; 13.74–23.10), and exfoliant/peel/scrub with burn‐related events (ROR 13.39; 9.68–18.53). Within‐cohort comparators attenuated effect sizes but preserved core signals (e.g., eye‐area × ocular symptoms within‐broad ROR 9.14; 6.68–12.49). Case‐check analyses showed low product‐name concentration across signals, supporting category‐level interpretation.

openFDA cosmetovigilance data can be leveraged to map facial‐skincare adverse event patterns and prioritize plausible safety signals. Eye‐area products showed especially strong ocular‐related signals, and leave‐on products (masks, retinoid/anti‐aging, serum/essence) showed coherent burn‐ and swelling‐related signals.