Background
Glaucoma, the second leading cause of blindness in the world [
1], is an optic neuropathy characterized by optic disc cupping, retinal nerve fiber layer (RNFL) thinning, and progressive vision loss that starts in peripheral visual fields. Glaucoma patients are often unaware that their visual field loss restricts peripheral object detection [
2‐
4] and real-world scene perception [
5]. Visual field loss severity is associated with higher motor vehicle crash rates [
6,
7], which are at least 4.2 times higher in patients diagnosed with moderate-to-severe glaucoma relative to healthy comparisons [
8]. Neurocognitive performance, by contrast, is a strong predictor of driving safety [
9,
10] in healthy drivers that remains understudied in glaucoma drivers [
8,
11]. Efforts to mitigate increased motor vehicle crash rates in glaucoma drivers will require understanding the interaction between visual field loss and neurocognitive performance in this at-risk population.
Driving simulators are a safe and externally valid alternative to on-road vehicles for studying driving behavior in at-risk populations [
12]. Although there are limitations to comparing driving performance across simulated and real-world environments, due in part to computational constraints on replicating the complexity of reality and the extent to which drivers suspend disbelief in a simulator, driving performance is comparable across platforms. For example, previous driving simulation studies show higher crash frequency in glaucoma drivers relative to healthy drivers [
13,
14] that is predicted by degree of visual field loss [
13], which is in accordance with previous on-road driving studies [
6‐
8,
15]. In a population-based study of state motor vehicle crash records [
6], Kwon and colleagues showed that binocular visual field loss in upper, lower, and left visual fields – but not right visual fields – were associated with higher crash rates. During a computer-based hazard perception task that induced simulated gaze-contingent visual field loss in healthy volunteers [
16], Glen and colleagues showed that hazard detection speed was slower during conditions with, relative to without, visual field loss, where significantly slower speeds were observed in superior, relative to inferior, visual field loss conditions. Together, these studies provide converging evidence across simulated and real-world platforms to suggest that patterns of visual field loss are associated with increased crash risk, which may be explained by an inability to rapidly detect upcoming roadway hazards.
Mechanisms of how glaucomatous visual field loss contributes to predictors of increased crash risk, such as poor vehicle control [
17], remain unclear. One possibility is that peripheral visual field loss causes poor lane boundary tracking and, by consequence, poor lane maintenance and vehicle control. In an on-road driving study, Bowers and colleagues found that subjective qualitative ratings of lane maintenance errors by a driving instructor were higher in glaucoma drivers with more restricted binocular visual fields [
11]. Useful field of view (UFOV) divided attention performance also predicted lane maintenance errors under concurrent cognitive loads (i.e. merging, curve taking). Critically, however, interactions between binocular visual field loss and divided attention on lane maintenance errors, as well as their association with global neurocognitive performance, were unexplored. To our knowledge, current studies have yet to evaluate how binocular visual field loss and global neurocognitive performance interact to predict quantitative measures of lane maintenance in the driving simulator environment. Prior studies that have included neurocognitive measures in their design relied on dementia screening tools to adjust for mental status in their models without describing statistical contributions of mental status to their reported results [
6,
8]. Thus, parsing out independent contributions of binocular visual field loss and pre-clinical variability in neurocognitive performance to predictors of increased crash risk, such as poor vehicle control, remain unexplored.
In the current pilot study, we sought to understand the relationship between binocular visual field loss, global neurocognitive performance, and quantitative measures of motor vehicle control (e.g. lane maintenance) in a driving simulator study of legally-licensed and actively driving glaucoma patients. To this end, we calculated binocular visual field parameters by integrating monocular Humphrey visual fields, assessed performance on the Montreal Cognitive Assessment (MoCA), and quantified both lateral and longitudinal vehicle control measures in glaucoma patients and suspects completing simulated driving scenarios. Given that vision precedes neurocognitive functions in models of the information processing pathway [
18], study results will be useful for determining whether binocular visual field loss explains independent or overlapping sources of variability in driving performance in glaucoma drivers.
Discussion
We studied the association between binocular visual field loss, global neurocognitive performance, and quantitative measures of simulated motor vehicle control in drivers diagnosed with glaucoma. Glaucoma drivers with a range of binocular visual field loss, relative to suspects without visual field loss, showed significantly worse lane position control as indexed by greater than a three-fold increase in lateral acceleration and steering wheel variability. Binocular visual field loss, neurocognitive performance, and self-reported driving performance were independently associated with measures of lane position control. Specifically, greater lateral acceleration variability was associated with: (1) greater binocular visual field loss; (2) worse neurocognitive performance; and (3) worse self-reported driving performance. Together, this work adds to our understanding of how driving behaviors, such as motor vehicle control, may contribute to motor vehicle crash risk in glaucoma by establishing a previously unexplored link between on-road driving results showing an association between binocular visual field loss and subjective qualitative ratings of lane maintenance errors [
11], on the one hand, and driving simulator [
13] and real-world driving [
6] results showing an association between binocular visual field loss and simulated crash risk, on the other hand. Furthermore, this work demonstrates non-overlapping contributions of binocular visual field loss and global neurocognitive performance to driving performance in glaucoma.
Previous driving simulator studies have demonstrated increased simulator crash risk in glaucoma drivers [
13,
14]. Szlyk and colleagues (2005) found increased simulated crash risk was associated with greater binocular visual field loss [
13], where 57% of glaucoma drivers with less than 10 degrees of total peripheral visual field were involved in simulator accidents. Using glasses to artificially constrict concentric vision, Udagawa and colleagues showed that the number of driving simulator accidents was significantly higher in drivers with vision constricted to 10 and 15 degrees of visual angle, where greater visual constriction was associated with higher crash rates with vehicles approaching from the visual periphery [
35]. Kwon and colleagues demonstrated increased crash risk in glaucoma drivers with binocular visual field loss in upper, lower, and left – but not right – visual fields [
6], though how visual field loss contributed to crashes remains to be determined. Together, these results demonstrate that peripheral vision loss impedes detection of peripheral roadway objects, delaying object processing and resulting in increased risk in failing to avoid a motor vehicle crash. Interestingly, however, glaucomatous visual field loss does not delay detection of medium-to-high contrast peripheral visual stimuli during a driving simulator divided attention task [
36]. By contrast, mechanisms explaining the relationship between peripheral visual field loss and predictors of increased crash risk, such as motor vehicle control, remain unknown. Drivers with severe peripheral visual field loss may be less able to track lane boundaries and maintain continuous control over lane position, which could explain increased simulator crash risk. In the current driving simulator study, we found greater binocular visual field loss was associated with greater lateral acceleration and steering wheel variability on a straight hazard-free highway. To further explore mechanisms explaining the putative relationship between poor motor vehicle control and increased crash risk in glaucoma, it will be necessary to demonstrate: (1) poor lane maintenance control mediates the relationship between binocular visual field loss and simulator crash risk; (2) greater binocular visual field loss is associated with perceptual deficits in lane boundary detection; and (3) improving lane boundary detection decreases simulator crash risk.
Driving simulators offer a safe environment for exploring mechanisms of driving safety risk. To evaluate their external validity, however, results from the current driving simulator study must be translated to on-road driving studies [
6‐
8,
15]. Wood and colleagues (2016) reported poorer subjective ratings of lane positioning in glaucoma driver relative to controls [
37], though the relationship between lane position and binocular visual fields was unexplored. Bowers and colleagues (2005), by contrast, found that more restricted binocular visual fields were associated with subjective ratings of on-road lane maintenance errors during curve taking [
11], suggesting our results may translate to on-road driving performance. Translating simulated and on-road driving performance, however, demands similar comparisons. Quantitatively evaluating on-road lane maintenance control measures, such as those reported here, has the additional benefit of automation to facilitate the rapid analysis of large driving datasets collected from instrumented vehicles equipped with remote sensors. To further explore the external validity of current driving simulator results to on-road driving performance and increased crash risk in glaucoma, further work is need to demonstrate: (1) greater binocular visual field loss is associated with quantitative measures of poor lane maintenance during an on-road drive in a remotely instrumented vehicle; (2) quantitative measures of poor on-road lane maintenance mediate the relationship between binocular visual field loss and motor vehicle crash rates as indexed by state records or long-term remote monitoring via real-world driving recorders; and (3) an association between simulated and on-road quantitative measures of poor lane maintenance.
Information processing models of driving performance are essential to understanding mechanisms of driving safety risk in medical populations [
38]. According to information processing models, neurocognitive functions are downstream of, and consequently impacted by, sensory function. Numerous studies have demonstrated how sensory dysfunction in glaucoma (i.e. peripheral visual field loss) contributes to driving performance and crash risk [
6‐
8,
15]. In healthy drivers, contributions of neurocognitive functions to driving safety risk are well-documented [
10,
39,
40]. To our knowledge, the current study is the first to explicitly evaluate the association between neurocognitive function and driving performance in glaucoma drivers. The handful of studies that have included neurocognitive measures in their study design relied on dementia screening tools that were utilized to statistically adjust for mental status without describing its contribution to driving performance [
6,
8]. Tatham and colleagues evaluated performance during a novel divided attention task that required detecting a variable contrast peripheral visual stimulus during a concurrent curve negotiation or car following driving task load in a simulator [
36]. Divided attention task performance was associated with concurrent car following or curve negotiation performance, visual field loss, and RNFL thickness, but not MoCA performance. Further, the authors found no difference in MoCA performance between glaucoma and control groups, which we replicated here, but associations between MoCA and driving performance were unexplored. In the current work, we found binocular visual field loss and global neurocognitive performance independently predicted simulated motor vehicle control performance, suggesting sensory and neurocognitive functions are both contributing factors to information processing models of driving safety risk in glaucoma. Indeed, emerging evidence shows that glaucoma patients are at risk for neurocognitive impairment [
41‐
43] that may be explained by brain network dysfunction in non-visual brain regions [
44,
45]. Accounting for neurocognitive performance in future studies will be important to further our understanding of increased crash risk in glaucoma.
Comparisons between simulated and on-road driving performance are limited by several factors, including the extent to which: (1) drivers suspend disbelief and immerse themselves in the simulated environment; and (2) the simulated environment replicates the complexity of on-road sensory experiences and roadway vehicle behaviors. Due to these factors, some drivers may be less likely to take driving simulations seriously, more likely to engage in more risky driving behavior, or not fully replicate their naturalistic driving behavior. Thus, driving simulator study results, such as those reported here, must be evaluated in the context of these limitations. Despite these limitations, driving simulators are a safe and externally valid alternative to on-road vehicles for studying driving behavior [
12]. Future applications of this research will benefit from exploring the current pattern of results during on-road driving performance.
The current study must be considered in the context of its limitations. (1) Study results were obtained from a relatively small sample size. Due to the limited study power, other factors and co-morbidities associated with driving (e.g. manual dexterity, fatigue) could not be explored. A larger study that replicates current study findings and includes additional factors will be necessary. (2) Glaucoma suspects were recruited as the control group because they present with the same pathophysiologic risk factors of glaucoma without visual field loss. A proportion of glaucoma suspects, including those enrolled in the current study, will progress to a glaucoma diagnosis with visual field loss. In future studies, enrolling an additional healthy control group will provide further support that findings reported here are related to visual field loss rather than glaucoma-related risk factors. (3) Glaucoma patients were statistically older than glaucoma suspects, suggesting key study findings may be interpreted as an age-related, rather than glaucoma-related, difference in simulated driving performance. Ruling out this alternative explanation, key findings remained when age was included as a separate factor in regression analyses. In future studies, it will be important to match comparison groups on age, among other demographic factors (4) Study results contribute to our mechanistic understanding of simulated driving performance in glaucoma drivers. To determine the external validity of these findings, driving simulator results demand translation to on-road driving performance. (5) MoCA was included in this study as a global measure of neurocognitive function. To better understanding which specific neurocognitive functions predict driving performance in glaucoma, more comprehensive neurocognitive assessments will be necessary in future studies.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.