This review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (Page et al., 2021) guidelines and was pre-registered on PROSPERO (ID: CRD42022370067).

A comprehensive literature search was conducted on 5th December 2022, using Academic Search Ultimate, CINHAL and MEDLINE, and Web of Science and Scopus. These databases, accessed through Lancaster University, were selected to address the multidisciplinary nature of the research question.

Two independent search strings, one pertaining to the population (MND) and one pertaining to eye movement tasks, of interest were developed. These two search strings were then combined with the ‘AND’ operator. The search strings applied to each database differed only in database specific dictionary terms included. Otherwise, the free-text search terms remained consistent. Appropriate free-text search terms were identified during scoping searches (see Appendix 1 for full search strings applied).

The sensitivity of our search strings was assessed through “search strategy testing” on each of the selected databases. To do so, we checked that our searches found three highly relevant papers that were identified during scoping searches. The following papers were used for this test: Sharma et al. (2012), Proudfoot et al. (2016), and Guo et al. (2022). The initially developed search strings proved to be sensitive (i.e., found all four papers), and so were not updated.

To ensure that relevant records published after the point of running the initial searches on 5th December 2022 were captured, we re-ran our searches on 30th November. To ensure that all relevant records were captured, we completed forward and backward citation tracking of included papers using Google Scholar. Moreover, to reduce the impact of publication bias on our review, we searched for relevant pre-prints using PsyArxiv, MedArxiv, and bioArxiv. Pre-prints were searched using combinations of the following search terms: (”motor neuron disease” OR “amyotrophic lateral sclerosis”) AND (”eye movements” OR “saccades”). Records that appeared relevant from the abstract were subjected to full-text screening.

The .ris files downloaded during the final search of each database were exported into CADIMA, an online open-access tool designed to facilitate each stage of conducting a systematic review (Kohl et al., 2018). The records were automatically de-duplicated using CADIMA’s de-duplication tool and were double-checked by hand. Of the 4,032 records identified, 817 (20.26%) of these were duplicates resulting in 3,215 articles eligible for screening. In circumstances in which we could not access the relevant record, Lancaster University Library requested access to these records.

Screening was conducted in two phases: (1) title and abstract screening and (2) full text screening. The criteria applied during title and abstract screening were: (1) the abstract is written in the English language, (2) this is an original piece of research, (3) the study involves human participants, (4) there is mention of oculomotor function or eye tracking technology (or related terms), and (5) there is mention of Motor Neuron(e) or related terms. The criteria applied to full text screening were: (1) the study involves a saccadic/smooth pursuit eye movement task, (2) a healthy control condition is included, (3) the study includes an MND group free from disclosed comorbidities, and (4) an experimental or pre-post design is used (no case studies).

It is extensively documented that various clinical conditions, including but not limited to neurodegenerative disorders [e.g., Alzheimer’s disease and Parkinson’s Disease (Anderson and MacAskill, 2013)], psychiatric disorders [e.g., anxiety disorders (Ainsworth and Garner, 2013), obsessive compulsive disorder (Bey et al., 2018), Schizophrenia (Bittencourt et al., 2013)], and neurodevelopmental disorders [e.g., Autism spectrum disorder (Caldani et al., 2020)], are associated with alternations in saccadic and smooth pursuit eye movements. The occurrence of comorbid psychiatric impairments is increasingly being recognized as prevalent in MND (Patten et al., 2007). Given psychiatric impairments can independently influence saccadic and smooth pursuit metrics individuals, papers that recruited only MND patients with disclosed comorbidities were excluded.

Two raters (MRR, MP or AD) independently screened all records during both stages of screening. The Inter-rater reliability of title and abstract screening was assessed and deemed to be “Good” (k = 0.69) by CADIMA for a small portion of records (10%) prior to screening any further records.

The level of inter-rate agreement for full text screening was also deemed to be “Good” (k = 0.75). When inconsistencies in rating arose, a third reviewer was involved in making the final decision. Of the 800 full-text records screened, 780 (97.5%) were excluded due to failure to meet our inclusion criteria (see Figure 2 for full breakdown for reason for exclusion). All records that passed the full-text screening phase were checked for retraction using the Retraction Watch Database (22 November 2023),¹ with none of the records being flagged as being retracted.

Data was extracted by a single reviewer (either MRR or MP) and was checked by the other reviewer to ensure no missing data and the data extracted was correct. The following data was extracted: participant groups; the number of participants per participant group; mean age of each participant group; diagnostic criterion applied; MND clinical subtype; MND age at onset; MND duration; MND site of onset; MND Progression Rate; MND severity; MND Survival; cognitive tests; eye tracking tasks employed; eye tracking device; direction of the main effects; any reported means and standard deviations (SDs); and any relevant conclusions made by the authors (see Supplementary Table 1 for data extraction table).

To provide an indication of the credence that should be given to individual studies we assessed the methodology quality of the papers that were analyzed. Methodological quality was assessed using the Mixed Methods Appraisal Tool (Hong et al., 2018), a tool which is suitable for evaluating methodological quality across heterogeneous designs. MRR and MP quality assessed all papers independently with excellent levels of agreement (95.7% agreement). Inconsistencies were resolved during discussions between both reviewers (see Supplementary Table 2 for full quality assessment).

A total of 23 papers met the inclusion criterion. All studies recruited an MND group and a control group free from any disclosed clinical condition that may affect eye movements. The majority (N = 11) of studies exclusively recruited people with ALS (Shaunak et al., 1995; Evdokimidis et al., 2002; Moss et al., 2012; Burrell et al., 2013; Witiuk et al., 2014; Gorges et al., 2015; Becker et al., 2019; Yunusova et al., 2019; Guo et al., 2022; Rekik et al., 2022; Raveh et al., 2023; Riek et al., 2023; Zaino et al., 2023), two studies recruited people with ALS and people with Primary Lateral Sclerosis (PLS) (Sharma et al., 2012; Proudfoot et al., 2016), two studies recruited people with ALS, progressive muscular atrophy (PMA) or PLS (Marti-Fàbregas and Roig, 1993; Poletti et al., 2021), three studies did not disclose the MND subtype (Abel et al., 1992, 1995; Donaghy et al., 2010), one study recruited people with spinal muscular atrophy (SMA) only (Anagnostou et al., 2021), one study recruited people with Kennedy’s disease (Anagnostou et al., 2019), and the final study recruited people with ALS and progressive bulbar palsy (Esteban et al., 1978).

In most studies (N = 10) MND was diagnosed in accordance with the Revised El Escorial criteria; however two studies applied the World Federation of Neurology diagnostic criterion, one study relied upon genetic testing, one study relied upon Electromyography and muscle biopsy analysis and the remaining six studies did not report the diagnostic criterion adhered to. Sample sizes in each study varied from as few as 7 to 864 people with MND (M = 77.86(182.2), Median = 32). A total of 1,670 people with MND (M = 61.85(123.48) and 1,419 controls were included across all studies. See Supplementary Table 1 for sample specifications.

The method of eye tracking applied substantially differed across studies with some (N = 2) employing bedside eye tracking techniques (i.e., clinical observation without an eye tracking device) and some relying on video/laptop camera recordings (N = 2). However, the majority of studies (N = 19) relied upon a video-based system that uses infrared light tracking video cameras (e.g., EyeSeeCam, Eyelink and Skalar Medical Iris Camera) (note the device was not recorded in one study).

Studies employed either one or a combination of the following tasks; (1) prosaccade task, (2) antisaccade tasks, (3) memory guided saccade task, and (4) smooth pursuit task. Successful completion of such tasks is thought to recruit distinct oculomotor pathways. Subsequently, following a general overview of the occurrence of oculomotor abnormalities the results obtained across the 21 studies will be presented by oculomotor task.

When considering the occurrence of oculomotor abnormalities, Poletti et al. (2021) concluded that oculomotor abnormalities could be detected at bedside examination in 10.5% of people with ALS studied. Moreover, employing videooculographic recording, Rekik et al. (2022) observed abnormalities in eye movement recordings in 72.6% of people with ALS.

When considering whether the occurrence of oculomotor abnormalities differs as a consequence of clinical presentation, Poletti et al. (2021) observed that oculomotor abnormalities were more common in people with bulbar onset compared to spinal onset illness. Moreover, Poletti et al. (2021) also observed that oculomotor abnormalities were more common with increasing disease severity (as quantified by the King’s staging system) and were more common in people with cognitive impairment. Interestingly, people who presented with oculomotor abnormalities typically were older at the point of ALS onset than people who did not present with oculomotor abnormalities (Poletti et al., 2021; Rekik et al., 2022). Thus, taken as a whole, Poletti et al. (2021) concluded that oculomotor atypicalities are a highly specific (94.5%), but less sensitive (35.0%), proxy for cognitive impairment in ALS.

Most studies (n = 19) employed the prosaccade task (see Table 1 for breakdown of observations by study). A substantial proportion of these studies observed that performance on the prosaccade task did not significantly differ between people with MND and healthy controls (Donaghy et al., 2010; Moss et al., 2012; Sharma et al., 2012; Burrell et al., 2013; Proudfoot et al., 2016; Anagnostou et al., 2019, 2021; Guo et al., 2022; Riek et al., 2023). The typicality of people with MND’s prosaccades was evident at the level of infrared tracking and video recordings. In particular, using video camera recordings Moss et al. (2012) observed that the number of vertical and horizontal prosaccade errors, the number of absent saccades (i.e., no movement when target provided), and number of anticipatory vertical and horizontal prosaccades made by people with ALS did not differ from that of controls.

However, some inconsistencies were noted. For example, Burrell et al. (2013), Proudfoot et al. (2016), and Guo et al. (2022) observed that prosaccade latency did not differ between people with ALS and controls. In contrast, Marti-Fàbregas and Roig (1993), Shaunak et al. (1995), Gorges et al. (2015), Rekik et al. (2022), and Zaino et al. (2023) observed that prosaccade latency was prolonged in people with MND (ALS, PLS and PMA patients) compared to controls. Alternatively, both Evdokimidis et al. (2002) and Witiuk et al. (2014) observed that people with ALS prosaccade latencies were significantly shorter than controls, with Witiuk et al. (2014) observing that the responses of many people with ALS fell within the express saccade epoch.

Concerning saccadic velocity, Shaunak et al. (1995), Burrell et al. (2013), Witiuk et al. (2014), Proudfoot et al. (2016), Anagnostou et al. (2021), Guo et al. (2022), and Riek et al. (2023) observed that peak velocity did not differ between people with SMA and ALS, and controls. In comparison Gorges et al. (2015), observed that peak velocities, especially in horizontal and upper directions, were significantly decreased in people with ALS compared to controls. Furthermore, Raveh et al. (2023) observed that peak velocity was increased in people with ALS compared to controls.

Atypicalities in prosaccades have frequently been observed in people with Alzheimer’s disease (Crawford et al., 2005, 2013; Anderson and MacAskill, 2013). Interestingly, Poletti et al. (2021) observed prosaccade dysfunction was significantly more common in people with ALS (occurring in 2% of the studied cases) compared to controls (p < 0.001). However, the frequency of prosaccade dysfunction did not differ between people with ALS and Alzheimer’s disease.

Importantly, whilst there are key differences in the MND phenotype sampled across studies (e.g., ALS vs PLS vs SMA), the discrepancies in observed results do not directly map to these differences in the MND phenotype. Therefore, it seems unlikely that the discrepancies across papers are due to differences in clinical populations sampled. Further supporting this assumption, there is a lack of clear consensus regarding the influence of clinical manifestation and disease severity on prosaccade metrics. For example, both Burrell et al. (2013) and Guo et al. (2022) observed that prosaccade parameters were comparable between people with bulbar and spinal onset. In comparison, Donaghy et al. (2010), Gorges et al. (2015), Raveh et al. (2023), and Zaino et al. (2023) observed that site of ALS onset significantly influenced prosaccade profile. Specifically, bulbar onset was associated with reduced prosaccade velocity (Zaino et al., 2023), increased prosaccade latency (Donaghy et al., 2010), increased prosaccade gain (Raveh et al., 2023), increased prosaccade duration (Zaino et al., 2023) and hypometric prosaccades in the upward direction (Gorges et al., 2015). In comparison, Zaino et al. (2023) observed that people with spinal onset produced vertical and horizontal prosaccades of reduced amplitude compared to bulbar onset patients. Furthermore, Esteban et al. (1978) observed that all people who displayed atypical vertical prosaccades had clinical evidence of severe pseudobulbar damage (quadriplegic). Concerning disease duration, Zaino et al. (2023) observed that the peak velocity of vertical prosaccades was inversely correlated with disease duration in people with bulbar onset, and Marti-Fàbregas and Roig (1993) observed that prosaccade latency decreased as the time from first symptom increased. This overall lack of clear consensus regarding the occurrence of prosaccade deficits in MND patients, and the influence of clinical manifestation and disease severity on prosaccade metrics may perhaps suggest that the prosaccade task lacks sufficient sensitivity to differentiate MND patients from healthy controls.

Fourteen studies employed the antisaccade task (see Table 2 for breakdown of observations by study). Consistent with prior literature considering alternative neurological conditions [e.g., Parkinson’s Disease (see Waldthaler et al., 2021, for review] and Alzheimer’s Disease (see Kahana Levy et al., 2018, for review), all but two studies, that measured error rate, demonstrated that the antisaccade error rate is significantly increased in both ALS and PLS compared to controls (Shaunak et al., 1995; Sharma et al., 2012; Witiuk et al., 2014; Gorges et al., 2015; Proudfoot et al., 2016; Becker et al., 2019; Yunusova et al., 2019; Rekik et al., 2022; Raveh et al., 2023; Zaino et al., 2023). In the two studies that did not observe elevated antisaccade error rates, Moss et al. (2012), and Riek et al. (2023) found antisaccade error rate to be comparable between people with ALS and controls. Considering the eye tracking techniques employed, Moss et al. (2012) analyzed antisaccades using bedside tracking techniques (video camera recordings). Thus, it may be that discrepancies between the observations of Moss et al. (2012) and all other antisaccade studies, that measured error rate, are attributable to differences in methodological sensitivity. It is, however, important to note that Yunusova et al. (2019) observed elevated antisaccade error rates in people with ALS when recording eye movements through a laptop camera. Moreover, Riek et al. (2023) observed comparable percentages of antisaccade direction errors when recording eye movements using an infrared light tracking device (Eyelink 1000); thus, it appears unlikely that this discrepancy is due to methodological facets.

Furthermore, a global increase in antisaccade latency was observed in people with ALS and PLS compared to controls (Shaunak et al., 1995; Donaghy et al., 2010; Sharma et al., 2012; Proudfoot et al., 2016). Importantly, Shaunak et al. (1995), Donaghy et al. (2010), Sharma et al. (2012), and Proudfoot et al. (2016) analyzed the latency of both correct and incorrect antisaccades as one variable. Interestingly, when analyzing the latency of correct and incorrect antisaccades independently, Evdokimidis et al. (2002) observed that correct antisaccades latency did not differ between people with ALS and controls, but the latency of error antisaccades was significantly shorter in people with ALS compared to controls. Thus, suggesting that deficits in inhibitory control (i.e., inhibiting the reflexive prosaccade response) may be influencing the latency of people with ALS’s antisaccades. Supporting this assumption, Witiuk et al. (2014) observed that the percentage of antisaccade errors made by people with ALS increased with increasing cognitive dysfunction, as indicated by decreasing Montreal cognitive assessment score. Moreover, degraded performance on the antisaccade task has been linked to volume reduction in the dorsolateral prefrontal cortex (Yunusova et al., 2019), a brain region frequently implicated in attention and working memory (Katsuki and Constantinidis, 2012; Barbey et al., 2013).

Considering the influence of clinical subtype on antisaccade metrics, both Sharma et al. (2012) and Proudfoot et al. (2016) observed that people with PLS made significantly more antisaccade errors and produced antisaccades of longer latencies than people with ALS, thus indicating that antisaccade metrics may be sensitive to differences in MND subtype. Moreover, within the ALS subtype, people with bulbar onset were both slower to initiate antisaccades (Proudfoot et al., 2016) and made more anti-saccade errors compared to those with limb onset (Proudfoot et al., 2016; Raveh et al., 2023). However, this observation was not consistent with Donaghy et al. (2010) and Zaino et al. (2023), both observing that there were no significant differences in antisaccade error rate between people with bulbar and spinal onset. Therefore, whilst antisaccade metrics may be sensitive to differences in MND subtype (e.g., ALS v PLS) they may not be sufficiently sensitive to more nuanced differences within each subtype (e.g., Bulbar v Spinal onset ALS).

There appears to be a lack of consensus regarding the influence of disease severity on antisaccade metrics. Specifically, Proudfoot et al. (2016) observed that disease severity, quantified by the ALSFRS-R score, but not disease duration or rate of progression, were related to antisaccade error rate. In comparison Zaino et al. (2023) observed that the number of correctly executed antisaccades was inversely correlated with disease duration.

Four studies employed the memory guided saccade task (see Table 3 for breakdown of observations by study). In doing so, it was observed that compared to controls, people with ALS made significantly fewer correct saccades (Gorges et al., 2015; Becker et al., 2019; Zaino et al., 2023) and made significantly fewer corrections of erroneous saccades (Zaino et al., 2023). While Evdokimidis et al. (2002) did not comment on the error rate on the memory guided saccade tasks in people with ALS compared to controls, Evdokimidis et al. (2002) observed that the latency of remembered saccades were significantly longer in ALS patients compared to controls.

Zaino et al. (2023) further explored whether specific clinical facets significantly influenced memory guided saccade metrics. In doing so, Zaino et al. (2023) observed that people with bulbar onset ALS made significantly more erroneous memory guided saccades, and in people with spinal onset illness, the percentage of correctly executed saccades positively correlated with parietal cortical gray matter volume. As only one study has sought to elucidate the role of clinical facets in memory guided saccade metrics, these observations should be treated with caution.

Ten studies employed a smooth pursuit task (see Table 4 for breakdown of observations by study). In line with prior literature which has consistently documented smooth pursuit abnormalities in people with alternative neurological conditions [e.g., Parkinson’s disease (Frei, 2021) and Alzheimer’s disease (Molitor et al., 2015), all bar one study observed substantial group level deficits in smooth pursuits in people with MND compared to controls (Abel et al., 1992, 1995; Marti-Fàbregas and Roig, 1993; Donaghy et al., 2010; Moss et al., 2012; Poletti et al., 2021; Guo et al., 2022; Rekik et al., 2022; Raveh et al., 2023)]. Poletti et al. (2021) and Rekik et al. (2022) observed smooth pursuit abnormalities in 6.9 and 56.5% of people with patients sampled, respectively. Interestingly, the one study, Gorges et al. (2015) that did not observe group level deficits in smooth pursuit eye movements, did however, observe atypicalities at the individual level.

When breaking down smooth pursuit metrics by type compared to controls, smooth pursuits in people with MND were characterized by increased frequency of saccadic intrusions (anticipatory saccades and square wave jerks) (Abel et al., 1992; Moss et al., 2012; Gorges et al., 2015) reduced gain (Abel et al., 1992, 1995; Marti-Fàbregas and Roig, 1993; Donaghy et al., 2010) and abnormal cogwheel eye movements (the interruption of smooth pursuit movements by catch-up saccades, resulting in jerky, uneven eye movement) (Gorges et al., 2015; Guo et al., 2022). It is, however, important to note that not all abnormalities were consistently reported. Specifically, whilst Abel et al. (1992), Marti-Fàbregas and Roig (1993), Abel et al. (1995), and Donaghy et al. (2010) observed reduced gain in people with MND (ALS and MND non-specified), Guo et al. (2022) reported that gain was comparable between people with ALS and controls.

When considering the influence of disease progression and severity on smooth pursuit metrics, Marti-Fàbregas and Roig (1993) observed that the number of saccadic intrusions increased with rate of progression. Moreover, Abel et al. (1992) observed that smooth pursuit gain was significantly more reduced, in response to increase in stimulus speed, in people with severe MND compared to people with moderate MND. Regarding the influence of clinical presentation on smooth pursuit metrics, Guo et al. (2022) observed that the occurrence of abnormal cogwheel smooth pursuit eye movements was more common in people with ALS who display bulbar involvement than people who do not display bulbar involvement. Comparably, Donaghy et al. (2010) observed that velocity gain was reduced to a greater extent in bulbar onset patients compared to spinal onset patients. Rekik et al. (2022) observed that the occurrence of abnormal smooth pursuits (i.e., smooth pursuits with a gain <0.8 at 3 Hz) was correlated with cognitive dysfunction, as classified by the Mini Mental State Examination (Folstein et al., 1983) and the Frontal Assessment Battery (Dubois et al., 2000). Some prior research (Morimoto et al., 2012) indicates people who present with bulbar-onset and bulbar involvement may be more susceptible to cognitive impairment. Thus, accumulatively the findings of Marti-Fàbregas and Roig (1993), Guo et al. (2022), and Rekik et al. (2022) may potentially indicate that the degree of bulbar involvement influences smooth pursuit deficits in people with ALS.