A simple headshot can reveal more than just your physical appearance. This study shows how tracking subtle changes in facial aging over time can help predict survival and reshape cancer treatment.
Research: Facial aging rates quantify changes in biological age and predict cancer outcomes. Image credit: hedgehog94/Shutterstock.com
Research published in nature communications To examine the ability of photo-based facial aging rate (FAR) to predict overall survival in cancer patients.
AI facial age as a measurable biological indicator
Biological aging rates vary widely among individuals and may influence cancer outcomes independent of chronological age. However, clinical use remains limited due to the lack of practical, non-invasive biomarkers that can be easily applied in routine care.
FaceAge is an artificial intelligence-based tool that estimates biological age from facial features such as skin texture, volume loss, and structural changes. Previous studies have shown poor survival outcomes in cancer patients predicted to be older than their chronological age, supporting its potential as a prognostic biomarker.
Measuring aging rate using Face Age
The authors previously developed a model called the Foundation Artificial Intelligence Model for Health Awareness (FAHR-FaceAge). The model was trained to recognize signs of poor health from over 40 million facial images. When used in conjunction with Face Age, patients whose estimated age was at least 5 years older than their chronological age had a 21% higher risk of death.
Based on this, the researchers examined serial photographs to understand signs related to disease progression and treatment response. Such long-term measures are already widely used in clinical practice. For example, changes in prostate-specific antigen (PSA) levels over time can help assess prostate cancer risk, and changes in blood pressure can provide insight into cardiovascular risk.
FAR and overall survival in cancer
Researchers conducted a retrospective study of 2,276 cancer patients receiving radiation therapy. Most participants were white, with a median age of 63.4 years, and 62.9% had metastatic cancer after their first radiation therapy course, increasing to 78.7% during their second course.
The researchers used two photographs of each patient taken as part of routine clinical practice for identification purposes at the beginning of each course of radiation therapy. These were used to predict biological age using the Face Age artificial intelligence algorithm.
FAR was calculated as the change in facial age divided by the time between photographs, providing a measure of aging rate. This was analyzed for correlation with overall survival.
Photographic intervals were categorized as short (10–365 days), medium (366–730 days), and long (731–1,460 days). Due to the small denominator, the range of FAR was much larger in the short-term group. Therefore, only FAR >20 was reported as significant in this group, whereas in the intermediate and long-term groups, the thresholds were set at FAR >10 and >1, respectively.
High FAR reduces overall survival
For many patients, their facial age was predicted to be older than their chronological age based on the second photo. Higher FAR was associated with poorer overall survival in all groups after adjusting for time between photos, gender, race, and cancer diagnosis at the second radiation treatment course.
The short-term group had a higher FAR and a 25% higher risk of death. In the intermediate and long-term groups, high FAR was associated with a 37% and 65% higher risk of death.
The researchers repeated the analysis only in patients with metastatic cancer. Although the same association was seen, there were more significant differences in survival outcomes between groups.
FAR is a stronger predictor of long-term survival outcome
They also examined the combined effect of the initial deviation of predicted facial age from chronological age (FADRT1) and FAR. This showed that patients always had the highest risk of death when both FADRT1 and FAR were high.
The longer the interval between photos, the smaller the difference in FAR values, especially in the long-term group. Yet, FAR remains a major predictor of survival outcomes, although it still plays an important role in increasing mortality risk.
This indicates that “FAR consistently outperforms FADRT1 as a prognostic marker across all time intervals, with the strongest performance in long-term intervals.”
Possible mechanisms underlying FAR-based predictions
The authors note that accelerated molecular aging, such as DNA damage and cellular senescence, often occurs at distinct tipping points, highlighting the nonlinearity of biological aging. For cancer patients, such dynamic parameters reflect not only the disease process but also the effects of cancer treatment.
By quantitatively measuring facial aging, FAR may be able to reflect changes in health status during the treatment period. Advantages of using FAR include accessibility, ease, cost-effectiveness, and repeated measurements to assess changes in health status over the course of treatment.
If validated, it could be incorporated into current prognostic parameters to identify high-risk patients across multiple cancer categories and guide decisions regarding monitoring intensity, supportive care, and treatment approaches, particularly in the setting of advanced disease where less intensive or palliative strategies may be appropriate.
Research limitations
The ethnic/racial and age composition of the sample limits the generalizability of the results. Additionally, the lack of data on disease progression and treatment meant that elevated FAR could not be interpreted as causal. Unmeasured factors such as cancer cachexia and treatment-related toxicity may have influenced the observed association between FAR and survival.
Because the photographs were taken at specific radiotherapy time points rather than at fixed intervals, the use of photographs may have introduced adaptation bias, as different interval groups may reflect different clinical scenarios and limit generalizability. Validation of this study is pending, and ethical and privacy concerns, and the potential for bias in such facial recognition systems, remain to be resolved before clinical translation.
Future research should correlate disease type, stage, and treatment in diverse populations using easily accessible algorithms with strong data protection barriers. The current findings need to be validated in prospective studies and in combination with other aging markers. If so, FAR could be a tool to support the delivery of personalized cancer treatment.
Click here to download your PDF copy.
