FaceAge predicts cancer survival using facial aging

Researchers at Mass General Brigham published results in Nature Communications reporting that an AI-derived facial age metric, FaceAge, and a longitudinal measure called Face Aging Rate (FAR) are associated with overall survival in patients with cancer. The investigators analyzed routine facial photographs from approximately 2,276 patients who received at least two courses of radiation therapy at Brigham and Women's Hospital between 2012 and 2023, according to ASCOpost and MedicalXpress. The study reports that higher FAR was linked to worse overall survival; ASCOpost cites adjusted hazard ratios of 1.25 (95% CI 1.03-1.51) for short inter-photo intervals (10-365 days), 1.37 (95% CI 1.00-1.86) for medium intervals (366-730 days), and 1.65 (95% CI 1.22-2.22) for longer intervals (731-1,460 days).

Per reporting in News-Medical and Medscape, FaceAge is derived from a model the authors call FAHR-FaceAge, which the teams describe as trained on large facial-image datasets reportedly exceeding 40 million images. The study computed FAR as the change in predicted FaceAge divided by elapsed time between photographs, using two routine identification photos taken at successive radiation therapy courses. The cohort was majority White with a median age around 63.4 years and a high proportion of patients with metastatic disease at baseline, as noted in the News-Medical summary.

AI estimation of biological age from images leverages visual correlates such as skin texture, soft-tissue volume loss, and structural facial changes, which models can learn as features correlated with chronological and physiological age. Industry-pattern observations: longitudinal biomarkers that quantify change per unit time, such as PSA kinetics in prostate cancer or serial imaging biomarkers, often add prognostic signal beyond single-timepoint measures. Applying the same idea to facial-appearance-derived biological age is a natural extension, but it faces distinct confounders such as lighting, camera angle, photo quality, baseline dermatologic conditions, and non-disease-related aging influences.

Editorial analysis: the study introduces FAR as a low-cost, noninvasive candidate prognostic biomarker that could complement existing clinical variables if externally validated. For practitioners, the appeal lies in extracting additional signal from routinely collected data (identification photos) without extra procedures. However, industry observers should note limitations reported or emphasized in coverage: the analysis is retrospective, drawn from a single health system, and the cohort demographics were skewed toward White patients, which raises concerns about generalizability and algorithmic fairness.

Editorial analysis: observers should look for independent external validation cohorts and prospective studies that:

• •reproduce the FAR-survival association across diverse populations and imaging conditions
• •compare FAR to established prognostic models using incremental predictive metrics (C-index, net reclassification)
• •test robustness to common confounders such as facial coverings, imaging device differences, and cosmetic interventions. Regulatory, ethical, and privacy discussions will also be relevant because use of facial images as health biomarkers intersects with biometric privacy and bias risk

ASCOpost and Euronews report co-senior author Raymond Mak, MD, saying, "Deriving a Face Aging Rate from multiple, routine facial photographs allows for near real-time tracking of an individual's health." ASCOpost also quotes Hugo Aerts, PhD, noting the potential for FaceAge tracking to inform personalized planning and follow-up.

What was reported in multiple outlets includes retrospective design, single-center data, and demographic imbalance in the cohort. Editorial analysis: those factors commonly limit immediate clinical translation for AI-derived biomarkers and necessitate explicit evaluation of performance across skin tones and clinical settings before deployment.

In sum, the study presents a reproducible analytic concept-quantifying facial aging velocity-as a candidate prognostic biomarker in oncology, supported by hazard-ratio associations reported in Nature Communications and summarized by major medical outlets. Editorial analysis: the result is hypothesis-generating and merits external validation, technical robustness checks, and careful ethical review before any operational use in clinical decision making.