Computer-based smile esthetic surveys based on slider technology allow more precise control of variables and the possibility of obtaining continuous data. Variations in the perception of smiles from different facial perspectives have not been resolved. The objective of this study was to quantify the ideal and the range of acceptable values for smile variables judged by laypersons from a full-face perspective for comparison with lower-face data.
Mirrored and symmetric male and female full faces previously determined by peers to be of average attractiveness were used. Ninety-six laypersons judged these smile variables: smile arc, buccal corridor fill, maxillary gingival display, maxillary midline to face, maxillary to mandibular midline discrepancy, overbite, central incisor gingival margin discrepancy, maxillary anterior gingival height discrepancy, incisal edge discrepancy, and cant. The judges manipulated the variables using adjustable image technology that allowed the variable to morph and appear continuous on a computer monitor. Medians for each smile variable were compiled, and the Fleiss-Cohen weighted kappa statistic was calculated to measure reliability. Multiple randomization tests with adjusted P values were used to compare these data with those for lower-face views.
Reliability ranged from 0.25 for ideal overbite to 0.60 for upper midline to face, except for upper and lower buccal corridor limits, which each had a kappa value near 0. There were no statistically significant differences between the ratings of male and female raters. The following variables showed statistically and clinically significant differences (>1 mm) when compared with the lower-face view: ideal smile arc, ideal buccal corridor, maximum gingival display, upper to lower midline, and occlusal cant. Although the smile arc values differed because of model lip curvature variations, the principle of tracking the curve of the lower lip was confirmed. For the full-face view, the raters preferred less maximum gingival display, less buccal corridor, more upper to lower midline discrepancy, and less cant of the occlusal plane.
Reliability was fair to moderate with the exception of the buccal corridor limits. Most variables showed no clinically meaningful differences from the lower-face view. The acceptable range was quite large for most variables. Detailed knowledge of the ideal values of the various variables is important and can be incorporated into orthodontic treatment to produce an optimal esthetic smile.
Smile esthetics has become a central concern for patients and orthodontists because this is a primary reason that patients seek orthodontic treatment, and orthodontists are now using this as a focus for treatment planning. Investigation of the variables that contribute to esthetic smiles began in a controlled manner with the innovative study of Kokich et al. Using altered photographs with only the lips and teeth visible to fabricate 5 variations of 8 variables, they asked participants to rate the attractiveness of the altered images on a visual analog scale (VAS). They found that laypersons, dentists, and orthodontists detected changes in smile characteristics at different threshold levels, and that laypersons were the most forgiving. This study began to define values for the smile variables. One drawback of the study of Kokich et al was the large increments they used to alter the images—in some cases, 2 mm between images. This made detection of small differences impossible and also left open the possibility that the true value for the variable was between the choices offered.
Johnston et al also showed a difference between orthodontists’ and laypersons’ ratings. These differences highlight the importance of focusing on what the patients want regarding orthodontic treatment, since they ultimately must be satisfied as long as their goals are within a clinically acceptable range.
The study of smile variables was advanced by using more sophisticated digital image manipulation and computer-based methodologies. Parekh et al studied smile arcs and buccal corridors, with raters viewing a series of incremental photographs with different combinations of ideal, decreased, and increased smile arcs and buccal corridors and made judgments regarding the ideal and the range of acceptable options for each variable. In the pilot study for this work, Parekh et al used a creative slider technology. This was a method of linking a slider to an oral image so that moving the slider altered selected portions of the image. The raters (all orthodontists) were asked to move the slider to choose the image representing the ideal smile.
Ker et al also used the slider method to study smile variables and were the first to use sliders for a full survey with lay raters. This technology allowed the raters to manipulate the variables themselves through a seamless range of possibilities and to choose the ideal and the acceptable limits instead of merely judging preselected images. The slider also was a change from the traditional use of a VAS to quantify esthetic judgments. A VAS is well established and considered valid and reliable. It is a subjective rating of the variable, and it is anchored to concrete concepts to make it valid. The slider allowed raters to view a large series of photographs quickly and choose the ideal easily. This was an efficient design that presented a wider range of possible choices in a shorter time and allowed a more precise selection.
Ker et al also used the lower-face perspective to compare their data with those of Parekh et al for oral image data while expanding the number of variables investigated. Ker et al looked at the following variables: buccal corridor fill, smile arc, maxillary anterior gingival height discrepancy, maxillary gingival display, incisal edge discrepancy, overbite, central incisor gingival margin discrepancy, canine torque in broad and narrow smiles, posterior crown torque in narrow and broad smiles, maxillary central incisor crown width to height ratio, maxillary lateral to central incisal ratio, maxillary midline to face, maxillary to mandibular midline discrepancy, and cant. Their study refined or defined the ideal for each of these variables and also a range within which the raters still considered the images to be acceptable. It is important to note the difference between ideal and acceptable. An acceptable range of values would be useful to clinicians in evaluating the smiles of their patients, especially for understanding that we cannot always achieve the ideal.
There are few studies of smile characteristics with a full-face perspective. The full-face perspective mimics views encountered in normal conversation in contrast to the lower-face and oral views. This wider perspective could dilute or de-emphasize the attention to the details of the smile. For example, a change in a variable will be much smaller relative to the overall image when viewing a full face rather than just the lower face. That appears to be the case as shown by the study of Flores-Mir et al, in which the esthetic impact of the anterior dental occlusion was less in the full-face view compared with the dental or lower-face views. This study also demonstrated significant variations by patient, most likely because of the model’s facial appearance.
Havens et al showed that photos of a malocclusion with a full-face view were more attractive than the same malocclusion shown as a circumoral view. Their theory was that the perspective of the face helped camouflage the unattractive oral area. Conversely, when Rodrigues et al showed people a series of photos with changes in smile arc, maxillary lateral incisor tip, midline diastema, and midline deviation, the perspective made no difference on their rankings.
The oral-view data of Parekh et al appeared similar to those of Moore et al with full-face perspective data for buccal corridors, but they were different from those of the lower-face view of Ker et al. Certainly, perspective has yielded contrasting results.
Shaw et al argued that overall facial attractiveness was more important than dental esthetics in overall esthetics. It is possible that the attractiveness of the face alters the importance of the smile characteristics and that the background attractiveness of the face must be accounted for and controlled so that this variable does not inadvertently bias the results.
The purpose of this study was to quantify smile variables from a layperson’s point of view with full-face images of models of average attractiveness. These data were compared with the same variables viewed from the lower-face perspective to determine the effect of perspective and further validate existing smile esthetics data.
Material and methods
The general method of this computer-based survey was to digitally modify 1 smile so that raters could evaluate the ideal and the acceptable range of several important smile characteristics, measured in the context of male and female full-face perspective images. The design was approved by the institutional review board of Ohio State University.
Raters were recruited from a poster displayed in a central campus facility. Those who were interested approached the investigators for more information; no raters were solicited. The raters were first given a script that briefly explained the study. Inclusion criteria required participants to be conversant in English and familiar with using a mouse to control a computer. They consented to participate by completing the study and providing optional demographic data (age and sex). Dental professionals and dentistry and dental hygiene students were excluded.
Photographs of faces of consenting young adults were digitally bisected, mirrored, and sized for the survey by using a photo editing program, Photoshop CS3 (version 10, Adobe, San Jose, Calif). These photographs were acquired from a database of facial images previously rated by peers to be of average attractiveness. This was accomplished by having young adults rate bisected and mirrored smiling frontal facial photographs of volunteer young adult models using a VAS scale anchored with “very attractive” and “very unattractive.” Faces with mean VAS values from the central 20% of the scale were used. Faces of average attractiveness were used to prevent any uncontrolled influence from the background attractiveness of the model. One female and 1 male face were selected.
A similar method was used to generate a set of symmetrical and esthetic teeth placed in the lip profile of the full-face images. An intraoral photograph of a completed orthodontic patient was bisected, mirrored, and reassembled to form a smile that was sized to fit the mouth by using Photoshop CS2. For each variable measured, sequential layers of the same smile were altered by using templates of teeth digitally separated from the initial image. Once a series of modification values was established that appeared to represent the range of visually realistic smiles, the tooth images were stored as sequences that showed small incremental changes in 1 variable that was suitable for combination with any of the facial images to create a finished stimulus model for rating. The variables examined in the study are described with the range of variations in Table I .
|Smile arc||The curvature formed by an imaginary line tangent to the incisal edges of the teeth, modified in varying degrees of curvature in relationship to the lower lip.||From no curvature to an accentuated curvature. The degree of curvature was in relation to the lower lip, so quantification differed for each model.|
|Buccal corridor fill||The amount of dark space displayed between the facial surfaces of the posterior teeth and the corners of the mouth, calculated as the total dark space on both sides of the mouth as a percentage of the total smile width.||From 6% to 26.5% in approximately 0.5% increments.|
|Maxillary gingival display or gummy smile||The amount of gingival show above the central incisor crowns and below the center of the upper lip. Negative numbers indicate gingival exposure; positive numbers indicate tooth overlap by the lip.||From 1 mm of gingival display (–1) to almost 7 mm of tooth coverage for the female model, and approximately 2 mm of gingival display (–2) to 6 mm tooth coverage for the male model in approximately 0.25-mm increments. The variation between models was due to differences in sizing and coordinating the images for different faces.|
|Maxillary midline to face||The relationship of the maxillary dental midline (measured between the central incisors) to the midline of the face, defined by the center of the philtrum and the facial midline.By definition, the ideal was considered to be 0 for this variable.||The maxillary midline was moved to the left of the face in approximately 0.25-mm increments. The right and left buccal corridors were maintained throughout the movement of the dentition.The maximum deviation shown was 6 mm.|
|Maxillary to mandibular midline||The relationship of the maxillary central to the central embrasure to the mandibular central to central embrasure.By definition, the ideal was considered to be 0 for this variable.||Maintaining the maxillary midline, the mandibular dentition was moved to the left in approximately 0.25-mm increments. The right and left buccal corridors were maintained throughout the movement of the mandibular dentition. The maximum deviation shown was 5.5 mm.|
|Overbite||The vertical overlap of the central incisors measured in both millimeters of coverage and percentage of coverage of the mandibular incisor. This was modified by incrementally altering the mandibular layer of the image in the vertical dimension. The vertical movement of the mandibular layer produced an increased or a decreased overbite.||The layer was moved in approximately 0.25-mm increments. The range was from 0 to 9 mm of overbite (or 100%).|
|Central incisor gingival margin discrepancy||The vertical gingival margin difference between the central incisors. By definition, the ideal was considered to be 0 for this variable.||The gingival margin of the left maxillary central incisor was altered in approximately 0.25-mm increments. The incisal edges were maintained at their original height. The maximum deviation was 3 mm.|
|Maxillary anterior gingival height discrepancy from central to lateral incisor||The difference in the vertical height of the gingival zenith of the central incisor to the lateral incisor. A negative value indicated that the lateral incisor gingival margin was incisal to the central incisor gingival margin; a positive value indicated that the lateral incisor gingival margin was apical to the central incisor.||Variations from increased to decreased height were presented in approximately 0.25-mm increments.The range was −2.6 to almost 1 mm.|
|Incisal edge discrepancy or lateral step||The vertical difference between the incisal edges of the central and lateral incisors.||Variation was assessed by moving both lateral incisors up or down together in approximately 0.25-mm increments. The range was 0.4 to 2.4 mm.|
|Cant||The divergence of the occlusal plane from the horizontal axis, as seen when smiling, was altered by gradually rotating the plane through a point between the central incisors. By definition, the ideal was considered to be 0 for this variable.||The rotation of the plane occurred in 0.25° increments. The range was 0° to 6°.|
The following were the dependent variables in this study.
Esthetic attractiveness of each variable: the perception of esthetics was based on the raters’ response to the instruction, “please adjust the slider below to the ideal image.” Smile characteristics could be adjusted by positioning a slider to a rater-determined ideal position. Each image was assigned a known value based on the deviation from the original image.
Acceptability: in separate images, the raters were then asked to select the position of the slider corresponding to increasing and decreasing the variable of interest relative to the ideal point identified by previous research. They were instructed to move the slider until the image became unattractive. By completing this exercise, they defined the limits of acceptability. Each image was assigned a known value based on the deviation from the original image.
Data were collected on a stand-alone laptop computer via a customized program running in MATLAB (Mathworks, Natick, Mass), a numeric computing environment and programming language software. The program randomly displayed a single face image with teeth and allowed the participant to use the mouse to adjust an on-screen slider according to the displayed instructions to choose the ideal image or the acceptable limit. The slider motion triggered changes in the tooth image displayed, allowing the participant to adjust through the full sequence of tooth images for 1 variable at a time. The increments were small enough between successive images to produce the illusion of continuous variation as the slider was moved. Every image for each variable had a number assigned to it that was identified by the program as the choice and saved as data by image number. The image numbers were translated to values that represented the modification value of that smile characteristic.
Of the 10 variables, 6 had 3 questions associated with them: choose the ideal image, the upper limit, and the lower limit. These were buccal corridor fill, smile arc, maxillary anterior gingival height discrepancy, maxillary gingival display, incisal edge discrepancy, and overbite. The other 4 had only 1 question: deviation from 0, because the ideal was defined as no deviation. These were central incisor gingival margin discrepancy, maxillary midline to face, maxillary to mandibular midline discrepancy, and cant. Each question was asked twice to assess the rater’s reliability.
To make the length of the survey manageable, the variables were divided into 6 surveys. Each variable was viewed completely by 1 group of raters (96 raters per variable according to the power analysis below). Surveys 1 through 4 included 2 variables and asked all questions for those variables. Surveys 5 and 6 included only 1 variable. It took most participants 10 to 15 minutes to complete 1 of these surveys. Each participant was compensated with a $10 gift card.
A power analysis was performed to determine the sample size. Of the dependent variables in this study, overbite was reported by Ker et al to have the highest variance, so it was used to determine the sample size. With a nondirectional alpha risk of 0.05 and assuming a standard deviation of 3.5, a sample size of 87 subjects was needed to detect a difference of ±1.5 mm with a power of 0.86. Ten percent was added to this sample size in case nonparametric analysis would be needed. As a result, the final sample size per variable was 96 subjects. With a sample size of 96 for each variable and 6 surveys, a total of 576 participants were required.
Median data were compiled, and a Fleiss-Cohen weighted kappa statistic (K W ) was used to confirm reliability. Multiple randomization tests with P values adjusted by using the step-down Bonferroni method of Holm were used to compare the data with those of Ker et al.
The raters were 51% male and 49% female. Their ages ranged from 18 to 72, with a mean age of 25.
The reliability statistics for our 10 variables ranged from 0.25 for ideal overbite to 0.60 for upper midline to face, except for the acceptable upper and lower limits of the buccal corridor, which both had a K W close to 0 ( Table II ).
|Measure||K W||LCI .95||UCI .95||Interpretation|
|Ideal smile arc||0.34||0.25||0.43||Fair|
|Maximum smile arc||0.31||0.22||0.41||Fair|
|Minimum smile arc||0.30||0.20||0.40||Fair|
|Ideal buccal corridor||0.36||0.26||0.45||Fair|
|Minimum buccal corridor||0.09||0.06||0.14||Slight|
|Maximum buccal corridor||0.03||−0.02||0.09||Slight|
|Ideal gingival display||0.49||0.41||0.56||Moderate|
|Minimum gingival display||0.58||0.52||0.64||Moderate|
|Maximum gingival display||0.46||0.38||0.55||Moderate|
|Upper midline to face||0.60||0.53||0.67||Moderate|
|Upper to lower midline||0.48||0.40||0.57||Moderate|
|Central to central gingiva||0.58||0.51||0.66||Moderate|
|Ideal central to lateral gingiva||0.35||0.25||0.44||Fair|
|Minimum central to lateral gingiva||0.48||0.40||0.55||Moderate|
|Maximum central to lateral gingiva||0.38||0.29||0.47||Fair|
|Ideal central to lateral step||0.30||0.21||0.39||Fair|
|Maximum central to lateral step||0.44||0.37||0.52||Moderate|