Do Patients with Glaucoma Have Difficulty Recognizing Faces?

Abstract

Purpose.: To compare face recognition performance of glaucomatous patients with age-similar visually healthy people.

Methods.: Percentage of correctly identified faces in the Cambridge Face Memory Test was assessed in glaucomatous patients (n = 54; mean age = 69) with a range of visual field (VF) defects and visually healthy controls (n = 41; mean age = 67). All participants underwent cognitive and visual assessment (binocular visual acuity [BVA], contrast sensitivity [CS], and Humphrey VFs, both 10-2 and 24-2) and had BVA of at least 0.18 logMAR. Patients were classified as having “early,” “moderate,” or “advanced” VF defects using the Hodapp, Parrish and Anderson method. Patients were also stratified by better-eye 10-2 mean deviation (MD) being better or worse than the 1% normative value.

Results.: There were no significant differences in age (P = 0.25) or cognitive ability (P = 0.31) between groups; however, differences in BVA and CS were statistically significant (P < 0.05). Patients with advanced VF defects identified fewer faces on average (±SD) (66% ± 15%) than those with early (75% ± 11%) and moderate (75% ± 13%) defects and controls (75% ± 11%); P < 0.05. Patients with a best-eye 10-2 MD P < 1% identified fewer faces (67% ± 13%) than those with 10-2 MD P > 1% (77% ± 11%) and controls P < 0.01 (75% ± 11%). Multiple regression analysis revealed CS was important for face recognition.

Conclusions.: When compared with age-similar people with healthy vision, glaucomatous patients with advanced bilateral 24-2 VF loss, significant 10-2 VF loss, or poor CS are more likely to experience problems with face recognition.

Introduction

Glaucoma is a chronic and progressive disease of the optic nerve that can ultimately lead to irreversible blindness. Whilst treatment can manage disease progression in most patients, any visual loss that has occurred may still affect day-to-day functioning. Research centering on responses to self-report questionnaires indicates that patients perceive difficulties carrying out certain activities, ^1–6 but less is known about how glaucoma affects the individual's actual performance in vision-based tasks, and at what stage of the disease. There is some evidence that patients with glaucoma have difficulties when carrying out tasks such as walking, ^7,8 balance, ⁹ reading, ¹⁰ driving, ¹¹ reaching and grasping, ¹² and searching for objects. ¹³ Another visual skill that may be affected by glaucoma is the ability to recognize faces. Faces are fundamental to our social behaviors, and our ability to recognize them is integral to many of our social judgments and relationships. ^14–16 Impairment in face recognition as a result of vision loss might therefore be expected to diminish a person's ability to participate efficiently in social interactions and to impact their quality of life (QoL). Patients with AMD have impaired face recognition ^17–19 and consider this disability significant in their everyday functioning. ²⁰ Furthermore, studies assessing the patient's visual ability across a range of everyday tasks indicate that face recognition may be a skill that is at risk of being compromised. ^21,22 Nevertheless, there has been no direct research quantifying the impact of glaucoma on face-recognition performance. By convention, glaucoma mainly affects peripheral vision, and this precept probably explains the lack of interest in face recognition, a task guided by central vision. Nevertheless, glaucoma often coincides with diminished visual acuity and contrast sensitivity (CS), ^23–25 which are thought to play a key role in observed difficulties with face recognition in AMD. ^18,26 Furthermore, it has been proposed that glaucoma may affect higher level cortical processing, which may compromise success in more complex visual activities ²⁷ such as face recognition, and this is supported by psychophysical evidence for some glaucomatous patients exhibiting difficulty with perception of global forms, even in early disease. ²⁸  

This study aims to investigate the impact of glaucoma on face-recognition ability by comparing the performance of glaucomatous individuals, with a range of visual field (VF) defects, with that of age-similar, visually healthy people on a validated test of face recognition. The study examines whether performance can be predicted by the type and severity of the VF defect, and whether participants' perceptions of their face-recognition ability relate to their actual performance. 

Methods

Participants and Visual Function Testing

Patients were recruited from Moorfields Eye Hospital Trust London and the Fight for Sight Optometry Clinic at City University London. All patients had been clinically diagnosed with either primary open angle or normal tension glaucoma and had no other eye disease. Visually healthy controls (of a similar age to the patients) were recruited from the University Optometry Clinic. To be included in the study, all participants were required to have a corrected binocular visual acuity (BVA) measured in logMAR (logMAR BVA) of at least 0.18 logMAR (Snellen equivalent of 6/9), as measured using an Early Treatment Diabetic Retinopathy Study (ETDRS) chart. Contrast sensitivity was measured in log units with a Pelli-Robson (PR) chart (PRlogCS). For patients, VFs (SITA standard 24-2 and SITA standard 10-2) were recorded in both eyes on a Humphrey Visual Field Analyzer (HFA, Carl Zeiss Meditec, Dublin, CA) on the same day as the face-recognition study. Patients were recruited to provide a wide range of 24-2 VF defect severity in their best eye. In this instance, the best eye is defined as that with the better VF mean deviation (MD) because VFs are the focus of this investigation. The better VF is likely more representative of the patient's overall VF status and will be a close estimate of the binocular VF. ²⁹ Patients were then classified into three groups according to the severity of their best eye VF defect (Early, Moderate, or Advanced) using the Hodapp, Parrish, and Anderson (HPA) classification; this summarizes the overall extent of damage using MD values, while also taking the proximity of the defect to fixation into account. ^30,31 SITA-Fast HFA (24-2) VFs were conducted in both eyes for the control participants to simply ensure they had no VF defects that would compromise their role as a control in the study; these values were not used for analysis. All participants were graded within “normal limits” as specified by the Oculus C-Quant (Oculus GmbH, Wetzlar, Germany)—a stray light measure that can be used as a marker for level of lens opacity. In addition, participants completed a test of cognitive ability (Middlesex Elderly Assessment of Mental Status Test [MEAMS; Pearson, London, UK]). All participants stated that they were in good general health, with no problems with mobility, pain, anxiety, or depression. Participants were not taking medication that could affect cognitive ability, such as antidepressant medication and beta-blocker medication (systemic and topical). 

The study received approval from a United Kingdom NHS (National Health Service) National Health Research Ethics Service committee. The study conformed to the Declaration of Helsinki, and all participants gave their informed written consent. All data were cleansed of personal information before being transferred to a secure computer database at the University. 

Procedure

Face-recognition performance was measured binocularly using the Cambridge Face Memory Test (CFMT), ³² a validated test of face-recognition performance created to detect developmental prosopagnosia (a specific impairment with faces), which has also been applied in other contexts, such as autism. ^33,34 The CFMT was displayed on a 22-inch monitor (Iiyama Vision Master PRO 514, Iiyama Corp., Tokyo, Japan; 1600 × 1200 pixels at 100 Hz). Participants were seated with their head mounted in a headrest fixed 60 cm from the monitor, subtending a visual angle of 36.9° horizontally and 28.1° vertically. All images were displayed at an average luminance of 4.29 cd/m² (SD, 1.16). On average, the images subtended 7.4° horizontally and 11.1° vertically for the viewing distance. The average half angle of the face was 3.7° (equivalent of 6.5 cm width half face) meaning that the size of the images were equivalent to viewing a face at a distance of 101 cm in real life; a realistic distance for face recognition to occur in the real world. All participants wore a trial frame with the appropriate refractive correction suitable for the viewing distance. 

The participant was introduced to six “target” faces, which were displayed at three different viewing angles (left, frontal, and right profile) for 3 seconds each. After each set of face presentations, the participant completed a forced-choice recognition trial whereby they were required to select the target face from a selection of three faces. Participants said the number corresponding to the target face (1, 2, or 3) aloud, and their response was keyed in by the experimenter. Between each set of presentations, the participant fixated on a cross at the center of the screen; this ensured all individuals were equally prepared and studied the faces from the same starting location. After a 20-second review stage, whereby the participant's memory of the faces was refreshed by viewing all six targets simultaneously, the participant completed a series of further forced-choice recognition trials whereby they were required to select the face they had seen before. Participants could take as long as they liked to decide as the study did not proceed to the next trial until they made their choice. Participants received the same instructions, and everyone completed a practice trial prior to beginning the study. This methodology was the same as the test from the original validation study. ³² However, it should be noted that the last set of trials in the original CFMT, consisting of faces masked with artificial Gaussian noise, was not used. (The purpose of these trials in the original CFMT was to examine reliance on special face-recognition mechanisms highlighted in previous research that were specific to the disorder prosopagnosia ³⁵ ; this was deemed irrelevant for the current study population.) All participants thus took part in a total of 51 recognition trials. Example images from the first stage of CFMT are shown in Figure 1. 

Figure 1.

 

Example images from the Cambridge Face Memory Test (CFMT).

Participants also completed a modified version of the Visual Activities Questionnaire. Participants were asked to respond to the statements as though wearing their glasses or contact lenses, indicating on a five-point scale (never, rarely, sometimes, often, or always) how often they experienced each specified vision-related problem. 

Analysis

Face-recognition performance was calculated as the percentage of the 51 trials where the target face was correctly identified. SD of percentage scores for participants in the original CFMT study was 11%. ³² A sample size of n = 40 controls was set to detect an effect size of 10% at a power of 0.80 (statistical significance = 0.05). For comparisons between each of the patient groups and the controls (taking account of unequal sample sizes), this meant a sample size of at least n = 13 was required for each grouping of patients. The power calculations used the pwr.t2n.test function in the pwr package in the open-source environment R (Version 2.8.0). ³⁸ Responses to the five key self-report statements were scored from 1 to 5 (1 = never; 5 = always). Scores for each statement were summed to yield an overall “total self-report” score for each participant, with higher scores indicating more perceived impairment. Additional variables considered were age, percentage score in MEAMS, logMAR BVA, and PRlogCS. 

One-way ANOVA was used to compare mean values for dependent variables in each of the four groups (early, moderate, advanced, and controls). Post hoc Tukey pairwise comparisons were used to establish distinct differences between subgroups. Multivariate linear regression was used to explore predictors of face-recognition performance. All analysis was conducted using SPSS 18 (IBM Corp., Somers, NY). 

Results

Forty-one individuals with healthy vision (mean age, 67; SD, 8 years) and 54 patients diagnosed with glaucoma (mean age, 69; SD, 7 years) took part in the study. For patients, the range of best-eye 24-2 MD was −22.6 to +1.5 dB (mean, −7.2; SD, 6.6 dB), and the range of best-eye MD for the 10-2 VF was −18.7 to +1.4 dB (mean, −7.3; SD, 5.8 dB). The SITA-Fast VFs of all control subjects for all eyes were classified as within normal limits (mean, 0.02; SD, 1.1 dB). Table 1 displays demographic information for the participants. 

Table 1.

 

Summary of Participant Demographics

	Patients (n = 54)	Controls (n = 41)
Age	Mean, 69 years; SD, 7	Mean, 67 years; SD, 8
Sex	26 Male, 28 Female (52%)	20 Male, 21 Female (51%)
Ethnicity	100% Western-European	100% Western-European
Best-eye 24-2 MD (dB)	Mean, −7.2; SD, 6.6	Mean, 0.02; SD, 1.1 (SITA Fast)
Worst-eye 24-2 MD (dB)	Mean, −12.3; SD, 7.0	Mean, −0.20; SD, 1.0 (SITA Fast)
Best-eye 10-2 MD (dB)	Mean, −7.3; SD, 5.8	N/A
Worst-eye 10-2 MD (dB)	Mean, −12.2; SD, 7.6	N/A
logMAR BVA	Mean, 0.06; SD, 0.11	Mean, −0.05; SD, 0.09
PRlogCS	Mean, 1.83; SD, 0.19	Mean, 1.95; SD, 0.00
% Score in MEAMS	Mean, 97.5; SD, 4.3	Mean, 97.0; SD, 8.3

 N/A, not available.

Primary Analysis

When stratified according to severity of VF defect in the best eye using the HPA system, 21 patients were classified as having “early,” 13 patients with “moderate,” and 20 patients with “advanced” VF defects in their better eye. There were no significant differences in age (P = 0.25) or percentage score in MEAMS (P = 0.31) between groups. However, differences in logMAR BVA and PRlogCS were statistically significant (P < 0.01). In the CFMT, the controls correctly identified 75% faces on average, with patients with “early” and “moderate” glaucomatous defects also averaging the same score (mean = 75%; SD = 11 and 13, respectively). Patients classed as having “advanced” VF defects identified fewer faces on average (66%; SD, 15%) than the other three groups (see Fig. 3A). Only the advanced group performed differently from the controls (F = 3.20, P = 0.03; 1-way ANOVA, with post hoc comparisons set at the 5% level). The 95% confidence interval for the difference in average score between the patients with advanced VF defects and controls was between 1% and 17%. Those in the “advanced” VF group reported more problems with face recognition in the questionnaire, although the difference was not quite statistically significant (P = 0.08). Table 2 gives summary measures for the main variables in each of the groups and the corresponding outcomes from the 1-way ANOVA analyses. 

Table 2.

 

Summary of Mean (SD) CFMT Scores and Other Study Variables for Each of the Groups Stratified according to Severity of 24-2 Defect in the Best Eye and Corresponding Outcomes from the 1-Way ANOVA Analyses

	Group				1-Way ANOVA
	Early VF Defect (N = 21)	Moderate VF Defect (N = 13)	Advanced VF Defect (N = 20)	Control (N = 41)	F	P
% Correct in CFMT	75 (11)	75 (13)	66 (15)	75 (11)	3.20	0.03
Total self-report score	8.8 (3.4)	8.2 (3.7)	9.7 (3.5)	7.5 (2.5)	2.34	0.08
Best-eye 24-2 MD (dB)	−2.7 (2.2)	−7.6 (2.6)	−11.4 (6.5)	—	19.98	<0.00001
Best-eye 10-2 MD (dB)	−3.1 (3.4)	−4.8 (4.0)	−11.6 (6.5)	—	16.82	<0.00001
ETDRS logMAR VA	0.03 (0.12)	0.09 (0.08)	0.06 (0.09)	−0.05 (0.09)	10.30	<0.00001
PRlogCS	1.86 (0.15)	1.85 (0.19)	1.80 (0.21)	1.95 (0.00)	6.00	0.001
Age	68 (7)	68 (9)	71 (7)	67 (8)	1.40	0.25
% Score in MEAMS	98.8 (2.1)	96.9 (5.2)	98.7 (2.0)	97.0 (8.3)	1.22	0.31

Figure 2.

 

Self-report face recognition questions.

Figure 3.

 

(A) Box plots displaying range of face recognition scores in patient groups with different disease stage and controls and P values for pairwise comparisons. Blue symbols depict scores of example patients with more peripheral VF defects, whilst red shapes show scores for selected patients with more central damage ([B] displays VFs for these patients). (B) VFs (24-2 and 10-2) from the best eye of example patients from each disease classification group. Symbols depict the VFs of the individual's corresponding face recognition score shown in (A).

Note that when patients were classified into the HPA severity groups according to their worst VF (the eye with the worse MD), there were no statistically significant differences in performance in the CFMT between groups (F = 0.1; P = 0.48). Likewise, the 1-way ANOVA for total self-report scores was not significant when patients were classified by the worst eye (F = 2.3; P = 0.09). 

Secondary Analyses

Figure 3B displays examples of 24-2 and 10-2 best-eye VFs of some of the best and worst performers in the CFMT by each VF group. When compared with Figure 3A, it can be seen that those who performed better in the task appear to have more peripheral VF damage and little apparent defect in their 10-2 VF, whilst those who performed worse in the test have more apparent central damage within the 10° VF. The HPA classification for VFs considers the presence of central defects in addition to severity in terms of MD, and so this observation suggests that it could be the state of the central VF in particular that could in some way be driving the likelihood of impairment in the task. Patients were, therefore, reclassified accordingly using their best-eye 10-2 VFs. More specifically, patients were stratified into two groups depending on whether they had a “significant” MD in the 10-2 best-eye VF; that is to say, the P value of the MD on the HFA chart was flagged as <1% (worse than the 1% normative value). Thus, 29 patients were classified as having significant defects in their 10° of VF with 25 patients classed as not having significant central damage (MD, P > 1%). It was found that patients with a significant defect (MD, P < 1%) in the 10-2 VF of their best eye identified fewer faces (mean, 67; SD, 13%) than those without significant central loss (mean, 77; SD, 11%) and controls (mean, 75; SD, 11%). The 1-way ANOVA showed that these differences were statistically significant (F = 5.99; P < 0.01). Those classed as having significant central defects also self-reported more problems with face recognition on average (F = 3.75; P = 0.04). Mean corrected logMAR BVA was 0.09 (Snellen equivalent 6/7; SD = 0.08) and 0.03 (Snellen equivalent 6/6; SD = 0.12) in the group with and without significant central 10° VF defects, respectively; and −0.05 (Snellen equivalent 6/5; SD = 0.09) in the control group. Average Pelli-Robson (PR) logCS values were worse for those with significant central defects (mean = 1.78; SD = 0.21) than those without (mean = 1.91; SD = 0.11) and control (mean = 1.95; SD = 0.00) groups. The 1-way ANOVA revealed that these mean values for logMAR BVA and logCS were significantly different (P < 0.01). The groups did not differ in age (F = 2.00; P = 0.14) or percentage score in the MEAMS (F = 0.48; P = 0.62). 

The impact of face recognition on binocular central vision was also investigated using another measure; the integrated visual field (IVF). The IVF has been used in the context of assessing the VF component of visual disability ^12,39–42 and simulates a binocular VF without the use of further testing. It has been suggested that the IVF could be even more useful for gaining an idea of visual function than monocular best-eye measurements alone. ²⁹ The IVF method involves “overlapping” the VF points by taking the best sensitivity at each corresponding point from the two eyes. In this instance, a central IVF was calculated using age-corrected total deviation (TD) values from just the four central values from the 24-2 VFs (points at 3°). These values were chosen to confirm whether central points of the VF may be more important for predicting face recognition. Likewise, an IVF for the TD points of the 10-2 VF was also calculated in light of the finding that significant defects in the 10-2 VF appear to be important for face recognition. As shown in Figure 4, when Spearman's rho coefficients were used to explore the relationship between performance in the CFMT and specified IVF central locations and other exploratory variables, the weakest correlation was for 24-2 MD in the best eye. Integrating the central four points of the 24-2 VF, however, provided a measure that appeared to improve the relationship with face-recognition performance; and the mean IVF for the 10-2 VF even more so. The highest correlation coefficient was for the association between CFMT score and PRlogCS. 

Figure 4.

 

Scatter plots showing the relationship between percentage of correctly identified faces in CFMT and (A) best-eye 24-2 MD; (B) IVF of central points of 24-2; (C) best-eye 10-2 MD; (D) mean IVF for the 10-2 VF; and (E) PRlogCS.

Table 3 displays results for a multiple regression analysis whereby mean percentage of correctly identified faces was entered as the dependent variable, and the following parameters were entered together as exploratory variables: best-eye 24-2 MD, best-eye 10-2 MD, mean IVF of four central 24-2 points, mean IVF for 10-2 VF, age, PRlogCS, logMAR BVA, and MEAMS percentage. Results showed that these variables accounted for 41% of the total variance (r ² = 0.41). The best predictor in this model was PRlogCS (P < 0.01). Stepwise multiple linear regression was subsequently conducted to determine the best model that could account for the highest proportion of variance in face-recognition scores. The stepwise multiple linear regression used a standard backward selection criteria, meaning variables could be removed from the model if they caused a reduction in the r ² value that had an associated P value greater than 0.1. The best model for explaining variation in performance was a combination of the PRlogCS and the IVF for the 10-2 VF, which accounted for 39% of the total variance (r ² = 0.39 [Table 4]). No other variables provided a significant contribution to the effect on the percentage of correctly identified faces. 

Table 3.

 

Results of Multiple Linear Regression of Best-Eye 24-2 MD, Best-Eye 10-2 MD, IVF for 10-2 VF, IVF for Central 24-2 Points, logMAR BVA, Log CS, Age, and Percentage Score in MEAMS on Percentage of Correctly Identified Faces in the Patient Group (n = 54)

Parameter	Estimate	SE	P Value
Best-eye 24-2 MD	0.14	0.46	0.76
Best-eye 10-2 MD	−0.33	0.53	0.54
IVF for 10-2 VF	0.90	0.59	0.13
IVF of central 24-2 points	0.14	0.46	0.76
PRlogCS	35.93	9.98	0.001
ETDRS logMAR BVA	11.29	16.11	0.49
Age	0.13	0.22	0.57
% Score in MEAMS	0.01	0.35	0.98

Table 4.

 

Best Predictive Model according to a Stepwise Linear Regression Using the Same Variables as in Table 3

Parameter	Estimate	SE	P Value	95% Confidence Interval
PRlogCS	32.70	8.99	0.001	(14.34, 49.08)
10-2 IVF	0.67	0.33	0.05	(0.12, 1.33)

 Excluded variables include 10-2 MD, 24-2 MD, IVF of central 24-2 points, age, logMAR VA, and MEAMS.

Discussion

Research into the types of disability faced by people with glaucoma is increasing but may still be underrepresented in comparison with other disabling chronic diseases. ⁴³ Existing research in glaucoma tends to focus on tasks facilitated by peripheral vision, such as mobility and driving; disability with regards to face recognition, a task more associated with central vision, has until now been overlooked. Face recognition is said to be an underestimated form of visual disability, ⁴⁴ and so it seems a worthy potential candidate to explore in patients with glaucoma. The results from this study suggest that impairments with recognizing faces could be another important outcome for certain patients with glaucoma. It has previously been suggested that glaucomatous damage may lead to impairment with global processing tasks such as recognizing faces even at an early disease stage. ²⁸ However, the current study found that whilst face recognition difficulties were apparent in glaucoma, it was only in patients with advanced visual field defects that there was a reduced average face-recognition performance when compared with people with good vision of similar age. Furthermore, those individuals with more advanced bilateral glaucomatous VF defects also self-reported slightly more problems with face recognition, although the association was not quite statistically significant. Assuming that the study had sufficient power to detect the effect, this weak trend could indicate that whilst patients with advanced VF loss may be aware of some disability, they may underestimate the true level of impairment that they display in terms of task performance. This finding coincides with previous research suggesting that patient perceptions of their vision do not necessarily match up with their actual level of functioning. ⁴⁵ These discrepancies may be related to the patient's disease expectations; for instance, patients with the same level of visual loss may perceive their functioning differently depending on how well their expectations match their current experiences. ⁴⁶ It is therefore worthwhile to consider both the patient's perceptions and measurements of their actual functioning when assessing the impact of glaucoma on task ability. 

Whilst there was an obvious shift in performance for more advanced patients, Figure 3A shows that there was still considerable variability in the scores of participants within each severity group. Since the HPA method, used to classify the patients, places some emphasis on defects to central vision, with any patient with at least one sensitivity value of 0 dB in the central points of their VF being automatically classified as “advanced,” it seemed plausible that defects to central vision were influential in determining the observed effects. This was supported by examining the visual field gray scales of the best and worst performing patients within each group (Fig. 3B). Further investigation implied that defects in more central VF (10° or less) may indeed be important for predicting the likelihood of face-recognition disability. For instance, those with significant central 10-2 defects in the best eye performed worse in the CFMT than those without significant central loss and controls. There was also some evidence suggesting the importance of considering the sensitivities of both VFs combined. For instance, mean IVF of the four central points of the 24-2 VF and mean IVF for the 10-2 VF points had an improved relationship with face recognition performance, again suggesting the importance of central binocular vision loss in predicting potential impairments with recognizing faces. These findings signify the importance of vigilance of VF points close to fixation in clinical care in addition to routine 24-2 VFs. When entered into stepwise multiple regression analysis, the IVF for the 10-2 VF points together with PRlogCS were the best explanatory variables for face-recognition performance. Poorer PRlogCS coincides with increased VF loss and is often symptomatic for glaucomatous patients with intact VA. ²³ The importance of contrast sensitivity in this task also supports previous research suggesting it plays an important role in successful face recognition ^26,47 and a variety of everyday tasks. ⁴⁸ In particular, evidence suggests that sensitivity loss at medium to low spatial frequencies can lead to problems with detecting faces, ⁴⁹ and since glaucoma leads to increased degradation of low spatial frequency sensitive pathways, ⁵⁰ it seems plausible that reduced contrast sensitivity contributed to the face recognition difficulties displayed by some of the patients. Nevertheless, some studies, using different experimental designs to the one reported here, disagree that contrast sensitivity is important for face recognition and suggest that loss of VA is more debilitating. ^17,18,44,51 All the same, the participants in our sample had relatively good acuity (6/9 or better), so it is possible that these results are underestimating the true impact of VA loss on face-recognition performance. It is also important to note that the central 10-2 IVF and PRlogCS accounted for less than half of the variation in facial-recognition performance (R ² = 39%). There were therefore undoubtedly a multitude of other uncontrolled social, psychological, and personal factors that also contributed to the results in addition to those measured here: impairment with face recognition alongside more advanced glaucoma is therefore by no means a certainty. This variation in task performance within groups was also observed in previous studies using performance-based measures to assess the types of visual disability experienced by glaucoma patients. ^10,12,13  

As in any case-control study, there are limitations. Whilst an underlying premise was to look at how glaucoma might influence “everyday” functioning, the black and white images arguably removed the study from a real-world context. Likewise, external features such as hair are absent in the images in the CFMT to ensure that face, and not feature, recognition is occurring. Yet, obviously, these features will play some role in real-life face recognition. Still, the CFMT is a validated test that has been used in a variety of clinical situations ^33,51,52 and was developed based on the strengths of other face-recognition tests that were already widely used in research. The test also simulates how faces would be learned in the real-world, by allowing participants to gradually acquire information about the faces from a variety of different viewpoints and build up more detailed representations in memory. Moreover, this study did not examine any learning effect with the CFMT, although both patients and controls performed the test in exactly the same conditions without any prior training. Furthermore, whilst the results of the MEAMS test indicate that the groups were of equal general cognitive ability, it is unlikely that this test was sufficient to pinpoint more subtle individual differences in cognitive skills such as general memory ability, and this could be considered a limitation of the results. Nevertheless, the results appear to indicate that performance becomes increasingly affected when considering more central VF locations (Fig. 4), which coincides with other research suggesting that central scotomas directly impair the ability to recognize faces, ^18,19 and indeed performance in a variety of other everyday tasks such as reading ^54,55 and reaching and grasping for objects. ⁵⁶ This study also did not investigate the contribution of defects to specific regions of the VF beyond points close to fixation; quantification of the potential influence of different regions of the VF, such as nasal or inferior/superior locations awaits future research. 

Face recognition deficits, as a disability, should not be underestimated; they can have socially debilitating consequences for those involved, causing anxiety, embarrassment, and the subsequent avoidance of social situations. ⁵³ Identification of the types of visual defects that are likely to lead to problems with recognizing faces presents an important opportunity for the improvement of current educational and rehabilitation strategies that may help improve patient QoL. For instance, alternative strategies for remembering faces such as those used in prosopagnosia ⁵⁷ may help those patients with more advanced, central VF loss or reduced contrast sensitivity better adapt to their condition. 

Conclusions

Glaucomatous patients with more advanced visual loss performed worse in a test of face recognition than patients with milder defects and people with healthy vision of a similar age. Central VF loss appeared to be important regarding impairment in face recognition, as patients with significant loss of vision in their central 10° of VF performed significantly worse than those with less severe central defects and controls. Contrast sensitivity was an important variable for explaining face-recognition performance. These findings contribute to a better understanding of the impact of glaucomatous visual loss on everyday functioning, by identifying face-recognition impairment as an additional challenge that could be faced by some patients with glaucoma. 

Acknowledgments

The authors thank Ryo Asaoka for his invaluable help with patient recruitment. 

Footnotes

 Supported as part of a program of work funded by unrestricted grants from the Special Trustees of Moorfields Eye Hospital, the Merck Investigator Studies Program, and an Independent Investigator Research Grant from Pfizer Inc. The authors acknowledge a proportion of their financial support from the Department of Health through the award made by the National Institute for Health Research to Moorfields Eye Hospital NHS Foundation Trust and UCL Institute of Ophthalmology for a Biomedical Research Centre for Ophthalmology. The views expressed in this publication are those of the authors and not necessarily those of the Department of Health.

Footnotes

 Disclosure: F.C. Glen, None; D.P. Crabb, None; N.D. Smith, None; R. Burton, None; D.F. Garway-Heath, None