Introduction

The appearance of the face typically changes in different dimensions and directions with age1,2. This phenomenon, known as allometry, is the reason why young faces are distinct from old faces3. Growth and alteration in craniofacial morphology have generated interest in many fields of science, especially physical anthropology4,5 and genetics6,7. Additionally, alternative applications of craniofacial morphology, such as forensic science, have received considerable attention8. In the field of Korean Oriental medicine, facial metrics are regarded as representative and reliable characteristics for diagnosing a person’s Sasang constitution9. Previously, we performed a two-stage genome-wide association study of facial morphological traits in two large Korean populations (1,926 samples from the Korea Medicine Data Center (KDC) cohort and 5,643 samples from the Ansan-Ansung cohort)6.

We speculated that the ageing-related trends in each facial metric might be important for human evolution and appearance. In this study, we compared 15 facial metrics – (a) BRH, brow ridge height; (b) IW, intercanthal width; (c) OW, outer canthal width; (d) RPFH, right palpebral fissure height; (e) LPFH, left palpebral fissure height; (f) NBH, nasal bridge height; (g) ULH, upper lip height; (h) RULT, right upper lip thickness; (i) LULT, left upper lip thickness; (j) SW, subnasal width; (k) PNL, profile nasal length; (l) FBW, facial base width; (m) NTH(V), nasal tip height (vertical); (n) NTP(H), nasal tip protrusion (horizontal); and (o) LFW, lower facial width – for each of four age groups (40 s, 50 s, 60 s and over 70 s) using the KDC and Ansan-Ansung cohorts as the replication set (Fig. 1). For example, the brow ridge is the nodule or crest of bone that is situated on the frontal bone of the skull10, forming the boundary between the forehead itself and the tops of the eye sockets. Normally, in humans, the ridges arch over each eye, offering mechanical protection11. Typically, the arches are more prominent in men than in women and vary between different ethnic groups12.

Figure 1
figure 1

Facial changes from age 40 years to age 70+ years. (a) BRH, brow ridge height; (b) IW, intercanthal width; (c) OW, outer canthal width; (d) RPFH, right palpebral fissure height; (e) LPFH, left palpebral fissure height; (f) NBH, nasal bridge height; (g) ULH, upper lip height; (h) RULT, right upper lip thickness; (i) LULT, left upper lip thickness; (j) SW, subnasal width; (k) PNL, profile nasal length; (l) FBW, facial base width; (m) NTH(V), nasal tip height; (n) NTP(H), nasal tip protrusion; and (o) LFW, lower facial width.

Full size image

Results

Gender differences

The distribution characteristics of 15 facial metrics are described in Table 1 and plotted in Fig. 1 by gender and age group in the KDC and Ansan-Ansung cohorts. The prominent gender differences were identified by the sex differences in the overall sample, as shown in Table 1, and the age distribution of sex differences, as shown in Fig. 1. Among the 15 facial metrics, six metrics (OW, NBH, ULH, SW, FBW, and LFW) appeared to be greater in males than in females in the overall sample comparison in both the KDC and Ansan-Ansung cohorts. In the age distribution of gender differences in Fig. 1, 12 metrics (BRH, OW, NBH, ULH, RPFH, LPFH, SW, PNL, FBW, NTH(V), NTP (H), LFW) showed gender differences. Interestingly, 4 metrics (RPFH, LPFH, RULT, LULT) were larger in females than in males.

Table 1 Means and standard deviations of facial metrics in KDC and Ansan-Ansung populations.
Full size table

Age group comparison

Age group differences were analysed by Student’s t-tests comparing the 40 s age group to the 50 s, 60 s and 70 s age groups. BRH, ULH, and NTH (v) are slightly increased in the 60 s age group in both cohorts and in both genders (Fig. 1a,g,m). OW, RULT, LULT, and NTP (H) were slightly decreased in the 60 s age group in both cohorts and in both genders (Fig. 1c,h,I,n). RPFH and LPFH showed a considerably decreased pattern in both cohorts and in both genders, and the pattern was prominent in females in both cohorts (Fig. 1d,e). Although FBW and LFW showed a considerably decreased pattern in both cohorts and in both genders, the pattern was prominent in males in both cohorts (Fig. 1l,o). PNL showed a slight increase in both cohorts, but only in males (Fig. 1k). IW, NBH, and SW seemed to have no clear pattern (Fig. 1b,f,j).

To understand in greater detail the effect of ageing on facial metrics, we constructed linear regression models with age and/or body mass index as covariates. We identified a significant linear increasing tendency of BRH and NTP and a decreasing tendency of OW, RPFH, LPFH, RULT, LULT, NTH, and NTP in both cohorts and in both sexes (Table 1).

Ethnic comparison of the human orbital region

Facial metrics in the orbital region, including PFH, IW, and OW, were compared between Koreans and other ethnicities (Fig. 2). RPFH and LPFH were relatively low among non-Asians relative to Asians (Fig. 2A). The Korean subjects had a large OW, ranking between those of Japanese and Indian individuals, whereas the IW was of Koreans was smaller than those of other Asian populations, such as Japanese and Chinese (Fig. 2B,C). Interestingly, the gender difference in IW was no more prominent in Koreans than in other ethnic groups.

Figure 2
figure 2

Ethnic comparison of the human orbital region. (A) Palpebral fissure height, (B) intercanthal width, and (C) outer canthal width. The data on ethnicities other than Korean were obtained from Vasanthakumar et al.14.

Full size image

Discussion

In this study, we analysed 15 facial metrics for gender differences and for ageing-related changes in two large Korean cohorts. Among these 15 metrics, 12 showed gender differences and tendencies toward age-related change. Increasing tendencies were observed in BRH, ULH, NTH, and PNL, and decreasing tendencies were observed in OW, RPFH, LPFH, RULT, LULT, NTP, FBW, and LFW. The decreasing tendencies were not changed even when adjusted for body mass index, indicating that some other structures under the skin (such as collagen and muscle) were reduced in the aged population.

In the orbital region, BRH showed a tendency to increase with age in both the KDC and Ansan-Ansung cohorts. The BRH is the nodule or crest of bone situated on the frontal bone of the skull10; this ridge separates the forehead itself from the tops of the eye sockets. Normally, in humans, the ridges arch over each eye, offering mechanical protection11. Typically, the arches are more prominent in men than in women and vary between different ethnic groups12.

“RPFH” and “LPFH” are abbreviations for the right and left palpebral fissure height. The palpebral fissure is the elliptic space between the medial and lateral canthi of the two open eyelids. There are many studies of palpebral fissure height; this metric is approximately 10 mm in adults and is smaller in East Asian populations than in white populations13,14. The gender differences in palpebral fissure height have been well summarized in reports on a South Indian population14. The bilateral orbital region, which is part of the upper face, acts as a key determinant of the perception of facial attractiveness, youthfulness and health14. Our comparison of palpebral fissure height implies that a very small PFH is a typical Korean facial characteristic (Fig. 2A). IW did not show any alteration with age, but OW showed a prominent decreasing pattern. The ethnic comparison of Fig. 2B,C showed that our study group had a shorter IW than Japanese or Chinese individuals, but the gender difference was no more prominent in Koreans than in other ethnic groups. In contrast, the OW of our study population was relatively large and was similar to that of the Japanese population. Therefore, the characteristics of the orbital region of Koreans are a small palpebral height and a wide canthal region. In the nasal area, the NTH slightly increased with age, and the NTP slightly decreased with age, but the NBH did not show any tendency to change. Additionally, the profile nasal length increased slightly with ageing only in males. The nasal tip is a challenging part of the nose for plastic surgeons because of variations in the anatomy of the lower lateral cartilage15. Therefore, our results help to clarify the direction of surgical correction for nasal tip deformities.

In the orthodontic region, the ULH slightly increased with ageing, and the RULT and LULT showed slight decreases with ageing. The upper lip is an important attractive point for the smile. According to Hulsey, the “smile is one of the most effective means by which people convey their emotions”16. The ULH is affected by the gingival margin of the upper central incisors and influences the attractiveness of a smile16. Generally, a shorter ULH indicates a higher smile line. On average, the smile line was found to be 1.5 mm higher in women than in men17; similarly, in the current study, the ULH was greater in males than in women, indicating a shorter smile line in women (see Table 1g). Although the concept of beauty has changed throughout the centuries, the thickness of the upper lip has always been a subject of interest and has importance in every culture18. These results may be beneficial to forensic anthropologists, plastic and reconstructive surgeons, and orthodontists.

Regarding overall facial width, both FBW and LFW slightly decreased with ageing, and the tendency was more prominent in male than female subjects. Facial width is reportedly associated with testosterone levels in males19. Therefore, the decrease in testosterone in elderly males may decrease the facial width.

Conclusion

Our study found that 12 of 15 facial metrics changed with age in subjects of both genders. The changes in eye size, nose length, upper lip thickness, and facial width might be general trends in Koreans because of the consistent tendencies in two large independent cohorts. Changes in facial metrics could serve as indicators to predict age from photographs and would be helpful for aesthetic facial care. To our knowledge, there are very few studies on the alteration of facial metrics with ageing, particularly in groups over 40 years old. Moreover, our study did not analyse the facial ageing of other ethnic populations. Therefore, we hope that the currents results will be validated in other ethnic groups and applied to diverse fields.

Materials and Methods

Study participants

A total of 1,926 participants in the KDC cohort were recruited from 19 sites (Korean Oriental Medical Clinics) between 2007 and 20106, and a total of 5,643 participants in the Ansan-Ansung cohorts were recruited from two regions in southern Korea from 2009 to 2012 for the Korean Genome and Epidemiology Study (KoGES)6,20 (Table 2). The subjects were photographed with a neutral expression in both frontal and profile views under the following standard conditions: the hair should be pulled back with a hair band; the centre points of the two pupils should be horizontally aligned, as should the upper auricular perimeters; and a ruler should be placed approximately 10 mm below the chin to convert pixels into millimetres. All participants provided written informed consent to participate in the study. This protocol was approved by the Korea Institute of Oriental Medicine Institutional Review Board (I-0910/02-001), and all research was performed in accordance with the relevant guidelines/regulations.

Table 2 Study Population Characteristics.
Full size table

Craniofacial measurements

Detailed descriptions of candidate feature variables and the corresponding measurement methods have been provided in a previous report6,9. Briefly, the facial variables were limited to those that could be easily quantified. Facial feature points in frontal and lateral images were automatically extracted by detecting and analysing the face, eyes, nose, mouth, and contours via an in-house program in Visual Studio C++ using OpenCV (Open Source Computer Vision Library). The positions of the extracted points were confirmed by a well-trained operator (accuracy: 98.8% on average). Fifteen facial metrics, each defined by the distance between two facial points, were derived from the photographs by converting pixels into millimetres using MATLAB software21.

For statistical analysis, 5 severely skewed facial variables, including PNL, NTH(V), NTP(H), RULT, and LULT, were ln transformed. We removed the outliers, which were defined using the first and third quartiles and the interquartile range of each facial variable. In each facial variable, measurements below the first quartile – 2.0 × interquartile range or above the third quartile + 2.0 × interquartile range were defined as outliers and excluded.

Statistical analysis

For each facial variable, the mean length was compared between age groups (40 s vs 50 s, 60 s or 70+) using Student’s t-test. Additionally, the mean lengths of male and female facial variables were plotted (Fig. 1) for both the KDC and Ansan-Ansung cohorts. We used the following criteria to identify phenotypes that tended to change with age: the mean underwent a significant gradual increase or decrease (p < 0.05), and the pattern appeared to be similar between the KDC and Ansan-Ansung cohorts.