Detection of Strabismus by Non–Health Care Professionals in an Ethnically Diverse Set of Images

Importance  Understanding the criteria for when strabismus becomes detectable by non–health care professionals could influence the goals for determining the success of surgical intervention and how patients with such misalignments are counseled.

Objective  To examine the magnitude at which strabismus is detectable by lay observers in an ethnically diverse set of images.

Design, Setting, and Participants  Photographs of 12 ethnically diverse models (black, white, and Asian) were simulated to have strabismus from esotropia of 21 prism diopters (∆) to exotropia of 21∆. From July 1, 2007, to October, 1, 2008, images were presented to 120 non–health care professionals aged 21 years or older from the general community in Boston, Massachusetts, who were asked whether strabismus was present. Analysis was conducted from November 1, 2008, to March 31, 2009.

Main Outcomes and Measures  The threshold angle for detecting strabismus to enable 70% of lay observers to make a positive determination whether strabismus is present.

Results  In white and black models, the threshold allowing a 70% positive detection rate was higher for esotropia than for exotropia (P < .001 for both). For white models, the threshold was 23.2∆ (95% CI, 21.0∆ to 26.5∆) for esotropia and 13.5∆ (95% CI, 12.5∆ to 14.6∆) for exotropia. For black models, the threshold was 20.8∆ (95% CI, 19.2∆ to 22.2∆) for esotropia and 16.3∆ (95% CI, 15.5∆ to 17.2∆) for exotropia. Asian models showed an opposite trend, with the threshold allowing a 70% positive detection rate for esotropia (14.3∆; 95% CI, 13.2∆ to 15.7∆) being lower than that for exotropia (20.9∆; 95% CI, 18.0∆ to 24.6∆) (P < .001).

Conclusions and Relevance  Esotropia was easier for lay observers to detect than exotropia in Asian models, and exotropia was easier to detect than esotropia in white and black models. This information should be considered when managing patients who have concerns about the social significance of their strabismus. Future studies should include diverse individuals and make an effort to account for individual factors that may alter the perception of strabismus.

Eye contact plays an integral role in our daily interactions by providing information and regulating personal communication.¹ Individuals with apparent ocular misalignment, such as strabismus, may be subject to psychosocial sequelae, including lower self-confidence and heightened social anxiety.^2-5 In addition, those with strabismus have reported being wrongly accused of cheating and being perceived as less sincere.² Others have found that individuals with strabismus were judged more negatively with respect to communication skills, intelligence, attentiveness, leadership ability, competence, and emotional stability.⁶ Strabismus also has been shown to affect a person’s employability as well as negatively influence the ability to find a romantic partner.^7-9 Therefore, strabismus has been shown to have an effect on an individual’s self-perception as well as how he or she is perceived by others.

It is presumed that these psychosocial consequences are influenced by whether the strabismus is cosmetically apparent. A better understanding of the criteria for when strabismus becomes visually detectable could influence how we counsel patients with strabismus, and, when appropriate, such an understanding may aid in considering the functional goals for determining the success of surgical intervention.

Despite the importance of defining visually detectable strabismus, the criteria remain ill defined. Some have suggested that a deviation of less than 15 prism diopters (∆) is considered small and cosmetically acceptable.¹⁰ Another criterion used to define success for strabismus surgery is a horizontal deviation of less than 10∆ without respect to direction.^11-19

Few studies have looked at the magnitude at which strabismus becomes visually detectable.^20-22 Two of these studies used just 1 white female model, while the third used 5 white models and 1 black model. Given the lack of diversity in these studies, the question arises of whether the criteria for visually detectable strabismus are the same across individuals, regardless of race/ethnicity. The aim of this study was to investigate the threshold for lay observers to detect horizontal strabismus in an ethnically diverse set of images.

Box Section Ref ID

Horizontal strabismus was simulated by altering photographs of models. Twelve models (4 black, 4 Asian, and 4 white; equal numbers of males and females in each ethnic group) with normal ocular alignment were photographed from a distance of 1 m. Photographs were taken of each model looking at the center of the camera lens (orthophoric photograph) and at points 3, 6, 9, 12, 15, 18, and 21 cm to the right and left of the camera lens. All points were clearly marked on a ruler perpendicular to the line of sight of the model, corresponding to 3, 6, 9, 12, 15, 18, and 21∆ of eccentric gaze, respectively.

Adobe Photoshop (Adobe Systems) was used to combine the orthophoric photograph from one eye with each of the eccentric gaze photographs from the other eye, simulating both esotropia and exotropia in 3∆ steps up to 21∆ for each of the 12 models (Figure 1). A total of 180 images were included in the study, consisting of 15 images from each model: orthophoria as well as esotropia and exotropia at 3, 6, 9, 12, 15, 18, and 21∆.

The study design was reviewed and approved by the New England College of Optometry Institutional Review Board and adhered to the tenets of the Declaration of Helsinki.²³ All non–health care professionals (hereafter called participants) were older than 21 years and were required to sign a written informed consent following an explanation of the nature and possible consequences of the study. All models were students enrolled at the New England College of Optometry and signed informed consent following an explanation of the study. As part of the informed consent, models gave the authors written permission to use the images for data collection purposes. It was explained that, although unlikely, it was not impossible that models could be recognized as data collection was conducted in the local community.

Participants were recruited from July 1, 2007, to October 1, 2008, and given a standard written set of instructions with a brief definition of strabismus: “An eye turn or more commonly ‘crossed eyes’ or ‘wall eyed’ is a misalignment of the eyes. Basically, one eye is looking in a different place than the other eye. The deviated eye can be pointed up, down, out or in. It can be the right eye or the left eye. It can be turned in a large amount or a very small amount.” Images were presented in a randomized order at a distance of 1 m to 120 adults who were not health care professionals. Participants were asked to decide whether strabismus was present. Judgment on direction or laterality of the strabismus was not required or recorded. Images were presented one at a time and viewed only once. Each participant was shown 36 images consisting of 3 images from each model, with no participant viewing the same image twice. The images were randomized such that, for a given eye turn position (eg, 12∆) of an individual model, 24 participants were asked to respond whether strabismus was present.

For analyses of all models combined, the percentages of yes responses from esotropia of 21∆ to exotropia of 21∆ (15 positions) were estimated by fitting a linear function of the magnitude of simulated strabismus for esotropia and exotropia positions separately using pooled response data from 12 models. The response threshold was the solution to the regression equation when the response rate is set at 70%. Seventy percent was selected before data collection to be consistent with previously published literature on this topic.^20-22 Bootstrap analysis was used to construct 95% CIs for the estimated parameters: thresholds allowing a 70% positive detection rate for both esotropia and exotropia positions. To model the normal variability in the estimated parameters, the bootstrap analysis was applied by resampling the number of yes answers for each position of each model 2000 times and obtaining a series of pseudo response data sets. Within each iteration, new parameters for a threshold allowing a 70% positive detection rate were computed based on the pooled pseudo yes counts of all 12 models using the algorithm described above. The same analytical methods were also applied to models grouped with respect to sex and race/ethnicity as well as to individual models. All analyses were planned before data collection.

Analysis was conducted from November 1, 2008, to March 31, 2009. The 95% CIs and P values were computed using the empirical distribution of the threshold parameters constructed from a series of pseudo data sets. The 95% CI for the response rate at the orthophoric position was constructed based on a binomial model.

For all models combined, the yes response rate increased as the magnitude of strabismus increased (Figure 2). Overall, exotropia was easier to detect than esotropia, with the threshold allowing a 70% positive detection rate of 19.4∆ for esotropia and 16.3∆ for exotropia (difference, –3.1∆; 95% CI, –4.4∆ to –1.9∆; P < .001). The yes response rate for the orthophoric position (false-positive response rate) was 25% (95% CI, 20.0%-30.0%). The thresholds allowing a 70% positive detection rate at positions 6∆ or greater were different from those at the orthophoric position for both esotropia and exotropia (P < .002 for all) (Table).

For the female models, the threshold allowing a 70% positive detection rate was similar for esotropia (18.1∆) and exotropia (17.3∆) (difference, –0.8∆; 95% CI, –2.5∆ to 0.8∆; P = .31). For the male models, the threshold allowing a 70% positive detection date was greater for esotropia (20.9∆) than for exotropia (15.2∆) (difference, –5.8∆; 95% CI, –8.4∆ to –3.7∆; P < .001) (Figure 3). The false-positive response rate was not significantly different between female and male models (20.8% and 29.2%, respectively; P = .10). The mean difference in the threshold allowing a 70% positive detection rate for esotropia between males and females was 2.8∆ (95% CI, 0.4∆ to 5.5∆; P = .01), with females having a lower threshold. In comparison, the difference between males and females in the threshold for exotropia was −2.2∆, (95% CI, −3.8∆ to −0.8∆; P = .004), with females having a higher threshold.

For the Asian models, esotropia was easier to detect than exotropia, with a threshold allowing a 70% positive detection rate of 14.3∆ for esotropia and 20.9∆ for exotropia (difference, 6.6∆; 95% CI, 3.5∆ to 10.7∆; P < .001). For the black and white models, exotropia was easier to detect than esotropia (difference, –4.5∆; 95% CI, –6.6∆ to –2.7∆; P < .001 and difference, –9.6∆; 95% CI, –13.7∆ to –6.7∆; P < .001, respectively) (Figure 4). For black models, the threshold allowing a 70% positive detection rate was 20.8∆ for esotropia and 16.3∆ for exotropia. For white models, the threshold was 23.2∆ for esotropia and 13.5∆ for exotropia.

For esotropia, the threshold allowing a 70% positive detection rate was higher for white and black models than for Asian models (P < .001 for black vs Asian and white vs Asian; P = .16 for black vs white). For exotropia, the opposite trend was noted. The threshold allowing a 70% positive detection rate was highest for Asian models, followed by black models, and lowest for white models (all P < .001 for pairwise comparisons). The false-positive response rate was the highest among Asian models (36 of 96 [37.5%]), followed by white models (24 of 96 [25.0%]), and was the lowest in black models (12 of 96 [12.5%]) (P = .02 for white vs black; P = .06 for Asian vs white; and P < .01 for Asian vs black).

The plots for individual models varied considerably in terms of the response curve slope and threshold allowing a 70% positive detection rate. Moreover, some individual model curves showed flat, discontinuous, or even paradoxically declining response rates when the magnitude of strabismus increased. This variability may be owing to a sampling error and inter-participant variability caused by independent participants judging the presence of strabismus at different magnitudes.

This report examines the ability of lay people to detect horizontal strabismus and, specifically, the magnitude of strabismus that was positively detected 70% of the time. For all models combined, the threshold allowing a 70% positive detection rate was lower for exotropia than for esotropia, although the difference between exotropia (16.3∆) and esotropia (19.4∆) was only of borderline clinical significance. The threshold allowing a 70% positive detection rate was equal regardless of the direction of strabismus for female models but greater for esotropia than for exotropia for male models. Differences across ethnic groups were also found. For Asian models, esotropia was easier to detect than exotropia, while in white and black models the opposite was true. Last, the false-positive response rate (the percentage of orthophoric images that were judged as having strabismus) was highest for Asian models, followed by white models, and was lowest for black models.

The strengths of this study include a standardized set of written instructions and use of a realistic set of photographic images to represent strabismus. In addition, this study used multiple models (male and female) of multiple ethnicities, including black, white, and Asian, making the results of this study more generalizable to a diverse population. Limitations of this study include lack of information on ethnicity and sex of the 120 non–health care professionals. In addition, participants were instructed to look for strabismus by examining individual still photographs, which is not representative of a real-life casual social setting. It is likely that individuals in a real-life casual social setting would be less likely to recognize strabismus than when they are specifically instructed to look for it. However, it is also possible that in a real-life situation when an individual is moving his or her eyes, the appearance of strabismus could be exaggerated in secondary positions of gaze, which could have the opposite effect and make detection of misalignment easier in some cases.

The threshold allowing a 70% positive detection rate for all models combined found in this study conflicts with previously published studies by Weissberg et al²⁰ and Reinecke et al,²¹ in which 1 white female model was used. Both studies reported not only lower thresholds to achieve a 70% positive detection rate for exotropia (approximately 8∆) and esotropia (approximately 14.5∆) but also a more dramatic difference between exotropia and esotropia. These differences may be secondary to our use of multiple, ethnically diverse sets of images. This possibility is further supported by the finding that there were opposing differences in the threshold allowing a 70% positive detection rate across ethnic groups. Thus, when the data were pooled for the entire set of images, the differences in the thresholds between esotropia and exotropia became more negligible. Another explanation for the overall lower detection threshold in other studies is that observers viewed the same face continually, whereas in our study observers encountered a variety of faces and were less likely to make comparative judgments.

On the other hand, our results are similar to those of Larson et al,²² who used a more diverse set of images, with 4 adults, 2 children, and an equal number of male and female models. That study found the significant threshold for lay observers to detect both exotropia and esotropia to be equal (15∆), which is more consistent with our findings.

We found that exotropia was easier to detect than esotropia for male models, but there was no difference in female models. This finding may be explained in part by sexual dimorphism of periocular measurements,²⁴ such as the tendency for larger interpupillary distances in men,^25,26 which may make exotropia appear more prominent. This finding contrasts with the 1 previous study²² that reported on the influence of the model’s sex. That study found no difference in sex and included 3 female and 3 male models, of which 2 were children and all but 1 was white, which makes direct comparison difficult.

The ethnic differences reported in this study cannot be compared with previous literature because, to our knowledge, none exists. The finding that Asian models had more easily detectable esotropia may be owing to characteristic facial features, such as epicanthal folds and flatter nasal bridge leading to greater coverage of the nasal sclera. It is likely that lay observers may rely on observing visible sclera nasally and temporally to aid in strabismus detection.

The effect of the observers’ ethnicity on their ability to judge whether models of different or similar ethnicities have strabismus is of interest. In other words, would a group of Asian observers judging Asian models perform similarly to the observers in this study? We did not record observer ethnicity, but a study is currently under way to investigate this variable.

It is important to consider the 25% false-positive response rate for all models combined. In other words, 1 of 4 images of models with straight eyes was perceived as misaligned. This rate may seem higher than some anticipate and is likely explained by the circumstances in which participants were primed to the task of searching for strabismus. The false-positive response rate in this study was actually lower than that reported in other similar studies.^20,22 Regardless, if they were not instructed to look for strabismus, we would expect lay observers to be less sensitive in their ability to detect strabismus. Thus, it is likely that the threshold for lay persons observing strabismus in a more casual setting would be higher. For this reason, the thresholds determined in this study could be considered a conservative estimate of visually detectable ocular misalignment.

Nonetheless, it is interesting that there were differences in the false-positive response rate among ethnicities. Another way to interpret this finding is to consider that many patients may appear to have ocular misalignment even though they do not and ethnicity may exaggerate (in the case of Asian individuals) or minimize (in the case of black individuals) that possibility. Our finding that Asian models had the highest false-positive response rate supports the clinical belief that pronounced epicanthal folds typically found in those of Asian descent may lead to the appearance of strabismus. It also supports the clinical axiom that Asian indiviudals are more likely to present with concerns of pseudoesotropia, especially in infancy.

The opposite is also true in that many patients may have strabismus but their eyes may appear aligned. We found that for all models combined, strabismus magnitude of less than 6∆ was judged no different than orthophoric photographs. This finding should give confidence to patients with small degrees of ocular misalignment that their strabismus is not noticeable.

An evidence-based understanding of what makes strabismus visually detectable could be a valuable tool for physicians when managing patients with cosmetic concerns about their strabismus and could be useful to strabismus surgeons when setting outcome measures for success of reconstructive surgery for strabismus. We found that ethnic diversity is another contributor to the perception of strabismus. Future studies should use a variety of models of varying sex, race/ethnicity, and facial features in an effort to obtain a meaningful understanding of what makes strabismus visually apparent.

Submitted for Publication: May 29, 2015; final revision received August 30, 2015; September 7, 2015.

Corresponding Author: Kimberley W. Chan, OD, Department of Ophthalmology, Boston Children’s Hospital, Harvard Medical School, 300 Longwood Ave, Boston, MA 02115 (kimberley.chan@childrens.harvard.edu).

Published Online: October 29, 2015. doi:10.1001/jamaophthalmol.2015.4082.

Author Contributions: Drs Chan and Weissberg had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Weissberg.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Chan.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Deng.

Obtained funding: Weissberg.

Administrative, technical, or material support: All authors.

Study supervision: Chan, Weissberg.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.

Funding/Support: This study was partially supported by grants R24 EY014817 and T35EY007149 from the National Eye Institute.

Role of the Funder/Sponsor: The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: Andrea Vosbikian, OD, assisted with data collection. She was compensated via a stipend provided through grant T35EY007149 from the National Eye Institute. We thank the model for granting permission to publish this information.