Introduction

In 1980, Borkan and Norris [1] found that perceived age, defined as how old a person looks to external evaluators [2], could reflect the underlying biological age, a measure based on physical and physiological parameters (e.g., forced expiratory volume, basal metabolic rate). Additionally, they found that people with higher biological ages (i.e., worse physiological status) died sooner than their chronologic age-matched counterparts with lower biological ages. Since measuring biological age requires extensive patient testing and considering that facial skin is the most accessible organ to evaluate [3], it could be inferred that perceived age could be used to predict mortality. Therefore, our review aims to find whether a high perceived age is a risk factor for overall mortality and comorbidities.

Methods

Studies were identified by searching PubMed, CINAHL and Embase from inception to July 27, 2020. MeSH (Medical Subject Headings) terms for “perceived age,” “facial aging,” “mortality,” “survival,” “heart disease,” “cancer,” “lower pulmonary disease,” “stroke,” “diabetes,” “Alzheimer’s disease,” “kidney disease,” “bone status,” and “hypertension” were used. See Table 1, which demonstrates the complete inquiry input.

Table 1 Complete inquiry, as input in databases.
Full size table

Studies were included if they (1) measured perceived age or isolated facial characteristics of old age, (2) measured mortality risk, mortality rate, or comorbidity outcomes, and (3) were in English. No particular time frame or publication status was considered. The study selection process was performed following the PRISMA Guidelines. This process, along with reasons for exclusion, is detailed in Fig. 1.

Fig. 1
figure 1

PRISMA Flowchart. Study selection process following the PRISMA guidelines

Full size image

Eligibility assessment was performed by one reviewer, starting with the title of the studies and followed by abstracts and full-text evaluations. If there were doubts in selecting an article, a second author reviewed the article according to the inclusion criteria and both reviewers came to a consensus for the final decision. Data extraction was performed by one reviewer. Risk of bias of the included studies was assessed using the ROBINS-I (Risk Of Bias In Non-Randomized Studies–of Interventions) tool of the Cochrane Library for nonrandomized studies. A summary and graph were created using RevMan 5.3 (Cochrane Collaboration), which allowed for bias stratification in several domains (Figs. 2 and 3).

Fig. 2
figure 2

Risk of Bias Graph. The figure shows the risk of bias across studies. (-) stands for high risk of bias, (?) stands for unclear risk of bias, and (+) stands for low risk of bias. This figure was created using RevMan 5.3

Full size image
Fig. 3
figure 3

Risk of Bias Summary. The figure shows each article’s risk of bias in the different components of the analysis. (-) stands for high risk of bias, (?) stands for unclear risk of bias, and (+) stands for low risk of bias. This figure was created using RevMan 5.3

Full size image

Results

The inquiry identified 977 studies, of which 11 fulfilled the inclusion criteria. Four more studies that fulfilled inclusion criteria were later added by searching the reference list of included studies and high impact journals in the related field for a total of 15 included studies (see Table 2). Four studies evaluated patients face to face, while 11 evaluated the patients with photographs. The included studies evaluated the association of perceived age with mortality (six studies), cardiovascular disease (four studies), chronic obstructive pulmonary disease (COPD) (three studies), cognitive function (one study), and bone mineral density (one study). All studies that included facial wrinkling or photoaging assessment along with their perceived age evaluations used either standardized evaluation criteria or photographic computerized assessments. The following paragraphs describe the most relevant information regarding each study’s outcome.

Table 2 Summary of included studies.
Full size table

Mortality

In 1982, by assessing patients from the Baltimore Longitudinal Study of the Gerontology Research Center, Borkan et al. [4] found that the majority of older-appearing men of ages 45–75 years were more likely to die at the end of 15 years’ follow-up (p < .001). Almost 30 years later, Schnohr et al. [5], using patients from the Copenhagen City Heart Study, found that men with no gray hair had a significantly lower mortality rate than the rest (p < .05). On the other hand, there was a significant correlation (p < .01) between half and complete arcus senilis and mortality in women. Facial wrinkle severity, however, was not found to be correlated with mortality.

A few years later, Christensen et al. [6] photographed a group of twins from the Longitudinal Study of Aging Danish Twins and found that among the pairs in which a twin had died 2 years after the first evaluation, the longest surviving twin’s perceived age was considered to be approximately 1.15 years younger. Additionally, the oldest-looking twin died first in 73% of cases in the subgroup in which perceived age differed by ≥ 2 years. Subsequently, Christensen and colleagues [7] found that 7-year mortality risk increased by 8–19% per standard deviation (SD) increase in perceived age (p < .001). Consistent with these findings, Gunn et al. [2] found a 17% increase in 7-year mortality risk and a 6% increase in 12-year mortality risk per SD increase in perceived age (p < .05). Additionally, Dykiert et al. [8] found a 51% increase in 7-year mortality risk per SD increase in perceived age in women (p < .02).

Cardiovascular Disease

In 1995, Schnohr et al. [9] found that completely gray hair increased the probability of myocardial infarction (MI) in men (p < .01). In addition, facial wrinkling was associated with an increased risk of MI in men ≤ 55 years (p < .05). Gunn et al. [10] found that women with the lowest cardiovascular risk looked > 2 years younger than women with higher risk (p < .002). Among cardiovascular risk components, blood pressure was found to have a significant positive correlation with perceived age (p = .004).

In 2012, Kido et al. [11] found that younger-looking patients had a lower age-dependent increase in carotid intima-media thickness (CIMT) (p < .0054). Additionally, CIMT had a negative impact on looking young (p < .05). Miyawaki et al. [12] found that facial pigmentation was significantly and positively associated with CIMT (p < .03) and brachial-ankle pulse wave velocity (p < .02). Interestingly, obesity-related parameters were significantly associated with pigmentation (p < .0003).

Chronic Obstructive Pulmonary Disease

In 1994, Lange and Schnohr [13] found that in current and past smokers, patients with higher wrinkle scores had a significantly lower percent of forced vital capacity exhaled in the first second (FEV1/FVC) than patients with lower wrinkle scores (p < .05). The association was stronger in women than in men. The authors also concluded that the presence of a significant interaction between smoking and wrinkling implied that patients with wrinkles were more susceptible to the effects of tobacco on lung function. However, the weak association did not allow for risk stratification of airflow obstruction by facial wrinkle assessment.

In 2006, Patel et al. [14] found that independently of cumulative tobacco exposure, facial wrinkling was strongly associated with the risk of airflow obstruction (p < .05). Facial wrinkling in smokers was also associated with an increased risk of COPD (p < .02) and with the presence and extension of emphysema on computed tomography (p≤.05). Contrastingly, O’Brien et al. [15] did not find the same association but instead found that facial wrinkling scores were significantly correlated with the diffusing capacity of lung for carbon monoxide (p < .05). In addition, O’Brien et al. [15] found that skin elasticity, as measured by the skin viscoelastic modulus of the forearm, was inversely and significantly correlated with FEV1/FVC (p = .001) and emphysema quantified from computed tomographic images (p < .001).

Cognitive Function

As part of their previous studies, Christensen et al. [7] also measured patients’ cognitive function with the Mini-Mental State Examination (MMSE) and found that perceived age had a significant inverse correlation with MMSE scores (p < .001). Recently, Umeda-Kameyama et al. [3] found that perceived age was more strongly correlated with MMSE scores than chronologic age in women (p < .00003). Additionally, perceived age showed a better correlation with the Vitality Index in the total population (p < .00000003) and in women (p < .0000000009) than did chronologic age.

Bone Mineral Density

In 2015, Nielsen et al. [16] used whole body pictures of women to assess perceived age based on the hypothesis that physicians might misjudge a patient’s age based on her posture since this reflects the severity of spinal osteoporosis. The authors found that an increased perceived age was significantly associated with a lower bone mineral density (p < .04).

Discussion

To our knowledge, the first article discussing the association between perceived age and mortality was published by Borkan et al. [4] in 1982. Their study initiated in 1958 with data collection, and patients were reexamined at 18-month intervals until the time of death. Analysis of covariance and multiple classification analysis demonstrated that patients who survived until 1977 were perceived as significantly younger, with an estimated age of 1.04 years lower on average than their chronologic age. This study’s critical bias was its age evaluation method, which was done face to face by a physician. This evaluation was subject to external cues that might have influenced the evaluator’s estimate [8].

Three other studies suffered from the same limitation: two by Schnohr et al. [5, 9] and one one by Lange and Schnor [13]. These three studies were performed before the year 2000 when digital photography was not widely available. Until 1998, 35-mm film photography was the standard method for documenting patients’ physical appearance [17].

Among all studies, Christensen et al. [6] were the first to use photography assessments. The studies that came after theirs all used photography; however, they sometimes differed in the way pictures were taken. Facial photography, particularly in plastic surgery, should follow some general recommendations to increase perceived skin detail. There are contrasting viewpoints on background color, with some advocating for the use of dark blue [18] and others for medium to light blue [19]. These blue shades seem like good complements for all skin tones, yet darker shades might diminish the three-dimensional quality of the picture [19]. Additionally, backgrounds should be free of folds and creases and composed of nonreflective materials such as matte paint, wallpaper, or cloth [20]. Except for Patel et al. [14], who took their pictures in the medical photography department of Addenbrooke’s Hospital of Cambridge University Hospitals, the articles included in this review described the background as “neutral” [6] or did not describe it at all [2, 3, 7, 8, 10,11,12, 15, 16].

Appropriate lighting is another crucial characteristic of good quality facial photography. It is recommended that two light sources be arranged at 45° angles and above the patient [21,22,23]. Also, multiple flash units and softboxes or umbrellas should be used as diffusers to eliminate shadows and provide an accurate depiction of facial redness and pigmentation [24]. The articles in this review did not describe in detail the lighting conditions they used for obtaining the patients’ pictures, with some only claiming to use “the same” or “standardized” lighting conditions for all patients [8, 15, 16] and others not describing them at all [2, 6, 7, 10]. Excluding Patel et al. [14], only three articles were more specific. One stated the use of a shadowless lamp [11], one used a 400–600 lux light source [3], and one used a device that provided a scattered light source [12]. Therefore, most of the articles lacked precision in describing their photography-taking technique, posing a substantial risk of bias for identifying facial aging cues.

Several studies have searched for the specific facial cues that influence perceived age. Nkengne et al. [25] found that the eyes and lips areas and skin color uniformity were the most critical characteristics influencing perceived age. Subsequently, their results were replicated by Kwart et al. [26] and expanded on by Forte et al. [27], who found that crow’s feet and lips’ vertical rhytides were some of the essential factors influencing perceived age in lateral and frontal pictures, respectively. The included studies that evaluated facial wrinkling always used lateral pictures of the patients’ temporal regions, either at 45° [10] or 60° [12] angles. Although some studies did not specify the angle from which the picture was taken [14, 15], periorbital skin wrinkling was evaluated, contributing to a more reliable assessment of patients’ perceived age. Of note, although O’Brien et al. [15] obtained photographs from patients’ perioral region, the authors did not specify if they included the evaluations in the facial wrinkling score.

In 1993, Sherertz and Hess [28] concluded that perceived age is a subjective estimation and that it might not be clinically useful as a marker, mainly because of the influence of environmental factors on skin characteristics. After this, it has been said that the association between perceived age and mortality risk might be a consequence of exposure to harmful factors [8]. However, since the study by Borkan et al. [4] in 1982, which evaluated perceived age as a mortality predictor, five more studies evaluated this association, with results pointing to the existence of a significant positive association between increased perceived age and mortality risk, even when controlling for factors that influence skin aging [2, 4, 6,7,8]. Only one of these studies looked for specific facial aging characteristics associated with mortality [5]. Interestingly, although the study found gray hair and arcus senilis to be mortality predictors in men and women, respectively, it did not find an association with facial wrinkling, which was the only facial skin aging characteristic evaluated [5]. Since the rest of the studies evaluating perceived age found either a significant difference between the estimated ages of the deceased and survivors [4] or an increased mortality rate [6] or mortality risk [2, 7, 8] in older-looking patients, the set of facial aging characteristics as a whole may be what drives the association with mortality, instead of an isolated component. Nevertheless, considering that the study by Schnohr et al. [5] was conducted more than 20 years ago, new prospective studies evaluating the association between specific facial skin aging characteristics and mortality risk are warranted.

Information was scarce regarding the association between perceived age and comorbidities. For cardiovascular diseases, the relationship with MI risk [9], carotid atherosclerosis [11, 12], and general cardiovascular risk [10] was studied. Schnohr et al. [9] found that gray hair and facial wrinkling were associated with a significant increase in probability of MI in men. The latter held only for patients younger than 55 years. Although this relationship might seem intuitive, the associations were significant even after including chronologic age and other age-related parameters as covariates. In another study, CIMT was significantly and positively associated with perceived age [11], while in a second study, it was associated, along with brachial-ankle pulse wave velocity, with facial pigmentation [12]. Additionally, Miyawaki et al. [12] found that facial pigmentation was also significantly and positively correlated with obesity measures, concluding that atherosclerosis indices were correlated with perceived age in women through increased fat.

Studies have demonstrated that a higher body mass index is associated with a decrease in perceived age in men, with a tendency for significance in women [29]. Furthermore, Guinot et al. [30] found that skin age scores were lower than chronologic age in overweight premenopausal women but not in overweight postmenopausal women or nonoverweight women regardless of menopausal status. The link between increased adipose tissue and perceived age might, therefore, be correlated with the effect of estrogens, whose levels have also been found to be inversely correlated with perceived age [31]. Menopausal status should be considered a potential confounder when establishing associations with perceived age or aging facial characteristics, such as in the study by Miyawaki et al. [12]

Gunn et al. [10] found that women with lower cardiovascular risk looked significantly younger than those with higher cardiovascular risk. Additionally, they found a significant positive correlation between blood pressure and perceived age in women. Previous studies have found a significant inverse association of microvascular skin function with cardiovascular risk [32] and blood pressure [33]. Therefore, the altered microvascular function in patients with high cardiovascular risk or essential hypertension may promote premature aging of specific areas of facial skin. Although the literature has mostly focused on histopathologic analyses of aging skin [34, 35], functional evaluation of the microvasculature and its repercussions on aging characteristics is still lacking.

No author studied the direct association of COPD with perceived age. Instead, due to its relationship with smoking, facial wrinkling was studied. The first study done on this matter by Lange and Schnohr [13] initially used a group of patients containing nonsmokers and found a significant inverse association between facial wrinkle severity and spirometry measures related to COPD. However, after group stratification, the association disappeared for lifetime nonsmokers and remained significant for past and current smokers. Two studies identified a similar pattern of outcomes [14, 15], with one finding a significant positive association between facial wrinkle scores and extensive emphysema on computed tomography [14]. Therefore, the evidence included in our review suggest that facial wrinkling severity, measured by the Daniell scoring system, warrants systematic investigation as a COPD severity predictor.

In addition to evaluating the association between perceived age and mortality and in their attempt to prove that perceived age was a suitable aging biomarker, Christensen et al. [7] also found perceived age to have a significant inverse correlation with MMSE scores. The recent findings by Umeda-Kameyama et al. [3] corroborated this association, with the difference that their results only held for women. Additionally, Umeda-Kameyama and colleagues [3] also found a significant inverse correlation between perceived age and vitality measures. Despite the reduced number of studies, the results demonstrate that perceived age is a good predictor of cognitive decline. Lastly, as to bone mineral density, though the results of a study by Nielsen et al. [16] seem promising, studies following a similar methodology are required to confirm the association found by the authors.

A biomarker is defined as a specific analyte, anatomic feature, or physiological characteristic that is measured [36]. A more specific definition by the Institute of Medicine states that biomarkers are indicators of normal biological processes, pathogenic processes, or pharmacologic responses to an intervention [37]. Researchers have been looking for a reliable and reproducible biomarker of aging that can substitute chronological age for decades since not everyone ages at the same rate. As a result, a wide array of aging biomarkers have been proposed, primarily biochemical, such as interleukin-6 and other inflammatory cytokines [38], testosterone [39], mitochondrial DNA [40], and telomere length [41], among many others involved in metabolic processes [42]. The American Federation of Aging Research proposed a set of criteria specific for aging biomarkers. They stated that these biomarkers should (1) predict the rate of aging (i.e., identify where a person is in their lifespan), (2) monitor a primary mechanism underlying the aging process (not an effect of disease), (3) be able to be tested repeatedly without harming the person, and (4) be something that works in both humans and research animals, so that it can be tested in the laboratory before being validated in humans [43]. In this setting, perceived age might only fulfill criteria 2 and 3, serving to monitor the aging process without harming the evaluated person. Therefore, it would be inappropriate to consider it a proper biomarker of aging. However, its association with mortality and the previously outlined comorbidities warrant its consideration and further study as a risk factor for these outcomes.

Lastly, no individual factors have been identified to explain the association between facial aging and mortality or comorbidities. However, oxidative stress and genetic predisposition have been proposed as possible explanations [8]. The oxidative stress theory of aging states that tissue functionality is lost due to macromolecule damage generated by reactive oxygen species (ROS) [44]. In addition to apoptosis [45], ROS induce cellular senescence and decrease the cells’ replicating capacity [46, 47]. Although it is known that the protective response to ROS decreases with age [48], some people might be genetically predisposed to either increased ROS production or decreased scavenging, increasing the risk of organ damage and disease development [49,50,51]. Moreover, the basis of photoaging is the continuous exposure of the skin to UV radiation and environmental chemicals leading to ROS formation and skin damage [52,53,54]. Ultimately, it could be possible that by overwhelming the protective capacity of the skin to ROS in genetically susceptible individuals, environmental insults could make a person look older than they are. In short, looking older for one’s age might reflect an already defective ROS clearance, with more underlying organ damage compared to same-aged peers.

Conclusion

Our review demonstrates that perceived age promises to be a useful predictor of overall mortality and cardiovascular, pulmonary, cognitive, and osseous comorbidities. The relative absence of studies evaluating the association between perceived age and different comorbidities is a topic that must be addressed to support the eventual use of perceived age in the clinical setting. Additionally, our review also highlighted that authors do not always follow the photography recommendations that ensure optimal visualization of facial skin characteristics, and even if they do, they often do not adequately record it in their methodology.

Limitations

This study has several limitations. Although some studies specified asking patients not to use any hairstyle or facial products when taking the photographs [10], others did not. This poses a substantial bias, since using these products can artificially decrease a subject’s perceived age. Since only studies published in English were included in this review, some relevant studies may have been missed. Other limitations include the scarcity of studies reporting on this topic, the potential bias of misinterpreting data and results, and the study selection process, the latter being a potential source of bias common to systematic reviews.