Antisaccades: A probe into the dorsolateral prefrontal cortex in Alzheimer\'s disease. A critical review

Antisaccades: A Probe Into The Dorsolateral Prefrontal Cortex in Alzheimer’s Disease by

Liam Kaufman Simpkins

A thesis submitted in conformity with the requirements for the degree of Masters of Science Institute of Medical Science University of Toronto

© Copyright by Liam Kaufman Simpkins 2008

Antisaccades: A Probe Into The Dorsolateral Prefrontal Cortex in Alzheimer’s Disease Liam Kaufman Simpkins Masters of Science Institute of Medical Science University of Toronto 2008

Abstract The number of people living with Alzheimer’s Disease (AD) is projected to increase dramatically over the next few decades, making the search for treatments and tools to measure the progression of AD increasingly urgent. The antisaccade task, a hands- and language-free metric, may provide a functional index of the Dorsolateral Prefrontal Cortex (DLPFC), which is damaged in the later stages of AD. Patients with AD make significantly more antisaccade errors than controls, however, performance in mild AD has remained unexplored. We hypothesized that mild patients will make more errors than controls. Thirty AD patients and 31 age-match controls completed both laptop-based and clinical versions of the antisaccade task. Two thirds of patients with AD made significantly more errors and corrected less of their errors than age-matched controls. Our findings indicate that antisaccade impairments exist in mild AD, suggesting DLPFC pathology may be present earlier than suggested by previous studies.

ii

Acknowledgments The completion of my M.Sc. thesis would not have been possible without the generous help of others. I would like to thank my supervisor, Sandra Black, for her tireless help and constant feedback. Sandy has demanded the most out of me, which has helped to improve both my ability to write and give presentations, skills that will be transferrable to where ever I end up. Second I would like to thank both Jay Pratt and Brian Levine for providing helpful feedback that strengthened my thesis and ultimately my understanding of the subject matter. I would like to acknowledge the efforts of Jen Brae who helped to recruit many of the participants who were included in this study. The LC Campbell Cognitive Neurology Unit has been integral in many aspects of my thesis and its members have provided me with helpful feedback! Julian Kirk-Elleker, thank-you for your wonderful illustration of the laptop-based antisaccade task. Mark Chiew, thank-you for developing an easy to use antisaccade application for the laptop. Cori Atlin helped with video coding and with coding criteria. The LC Cognitive Neurology Unit, Ontario Graduate Scholarship and the Scace Graduate Fellowship (OSOTF) provided the funding for my graduate work, thank you! Participants were part of CIHR funded project MT13129 held by Dr. Black. Lastly, I would like to thank my family for their support of my work, my decisions and my enthusiasm for different types of science. To my parents, Michael and Maureen, your scholarship has helped to guide my decisions, thank you. Lisa, my wife and best-friend, thank you for your never-ending support, feedback and your ability to keep me going!

Liam

iii

Table of Contents List of Tables .................................................................................................vi List of Figures ................................................................................................vii List of Abbreviations.....................................................................................viii Chapter 1: Rational and Objectives ............................................................1 1.1 Rationale 1.2 Objectives 1.2.1 Determine if mild AD patients make more antisaccade errors than age-matched controls 1.2.2 Determine if more errors are made on trials preceded by a trial of a different direction relative to trials preceded by a trial of the same direction Figure 1 – Same vs. Different Trials

Chapter 2: Literature Review ......................................................................6 2.1 Alzheimer’s Disease Table 1 - Studies investigating antisaccade performance in Alzheimer’s Disease 2.2 The Antisaccade Task Figure 2 – Prosaccade And Antisaccade Tasks 2.3 Developmental Changes 2.4 Functional Imaging Studies 2.5 Focal Lesion Studies 2.6 Differentiating Alzheimer’s Disease From Healthy Aging Table 2 - Power Analysis and Differentiation Capabilities of the Antisaccade task in Alzheimer’s Disease 2.7 Antisaccade Errors and Dorsolateral Prefrontal Pathology in AD 2.8 Barriers in Adopting the Antisaccade Task as a Probe of DLPFC Function 2.9 The Antisaccade Task as an Index of Severity 2.10 Neuroimaging & Dementia 2.11 Fractionation of Processes in the Antisaccade Task 2.11.1 Fractioning the Antisaccade Task: Inhibition Control 2.11.2 Fractioning the Antisaccade Task: Fixation 2.11.3 Fractioning the Antisaccade Task: Vector Inversion 2.11.4 Fractioning the Antisaccade Task: Voluntary Saccades 2.12 Memory, Understanding & Attention 2.13 Task Sequence 2.14 Clinical Adaptation 2.15 Non-Alzheimer’s Dementia Table 3 - Studies investigating antisaccade performance in non-Alzheimer’s Dementia 2.16 Discussion 2.17 Conclusion

Chapter 3......................................................................................................33 3.1 Background 3.2 Hypotheses 3.3 Methods 3.3.1 Study Participants 3.3.2 Saccade Tasks

iv

Figure 4 – Experimental Setup 3.3.3 Saccade Coding 3.3.4 Statistical Analysis 3.4 Results Table 4 - Demographics Table 5 – AD Neuropsychology Results Figure 5 – Antisaccade Errors and Uncorrected Errors Table 7 - Diagnostic Capacity of Antisaccade Metrics Figure 6 – Same Direction Errors Compared To Different Direction Errors 3.5 Discussion 3.6 Clinical Antisaccade Task 3.7 Antisaccade Errors Elevated in Mild AD 3.8 Dementia Severity and Antisaccade Error Rates 3.9 Antisaccade Errors On Same Relative To Different Direction Trials 3.10 Prosaccade Errors 3.11 Antisaccade Errors and Neuropsychology

Chapter 4......................................................................................................57 4.1 Performance Monitoring and Task Setting Deficits 4.2 Clinical Relevance of The Antisaccade Task 4.3 Study Limitations 4.4 Future Directions 4.4.1 Fractionation: domain specific and domain general 4.4.2 Neural Correlates: Structural and Functional Imaging in Conjunction with Pathological Data 4.4.3 Large Scale Validation Of The Antisaccade Task In AD 4.4.4 Same Trials Versus Different Trials 4.5 Conclusions

References ....................................................................................................66

v

List of Tables 1. Studies investigating antisaccade performance in Alzheimer’s Disease – page 7 2. Power Analysis and Differentiation Capabilities of the Antisaccade task in Alzheimer’s Disease – page 16 3. Studies investigating antisaccade performance in non-Alzheimer’s Dementia – page 32 4. Demographics – page 44 5. AD Neuropsychology Results – page 44 6. Antisaccade Performance – page 46 7. Diagnostic Capacity of Antisaccade Metrics - page 47

vi

List of Figures 1. Same vs. Different Trials - page 5 2. Prosaccade And Antisaccade Tasks – page 9 3. Laptop Prosaccade and Antisaccade Tasks - page 38 4. Experimental Setup - page 39 5. Antisaccade Errors and Uncorrected Errors - page 45 6. Same Direction Errors Compared To Different Direction Errors - page 48

vii

List of Abbreviations AD – Alzheimer’s Disease AP - Amyloid Pathology BOLD - Blood Oxygen Level Dependent CBD - Corticobasal Syndrome DTI - Diffusion Tensor Imaging DLPFC – Dorsolateral Prefrontal Cortex EEG – Electroencephalogrophy FA - Fractional Anisotropy FEF – Frontal Eye Fields fMRI – Functional Magnetic Resonance Imaging FTD - Frontotemporal Degeneration MEG - Magnetoencephelography MMSE - Mini-mental Status Exam NFC - Neurofibrillary Changes PET – Positron Emission Tomography PSP - Progressive Supranuclear Palsy SEF – Supplementary Eye Fields WCST - Wisconsin Card Sorting Task

viii

1

Chapter 1: Rational and Objectives 1.1 Rationale Alzheimer’s disease (AD) is the number one cause of dementia and its prevalence is projected to increase three and four fold in developed and developing countries respectively (Ferri et al., 2005). New treatments and the potential for disease prevention have increased the need for early AD detection. In addition to early diagnosis, monitoring the progression of AD will be crucial in determining the effectiveness of various treatments. Recent technological advances coupled with increased volumes of data have led to a revised set of criteria for the diagnosis of AD that has placed a greater emphasis on the role of neuroimaging techniques (Dubois et al., 2007). For instance, MRI measured volumetric loss of the hippocampus surrounding structures, provides a sensitivity and specificity greater than 85%. Likewise, in vivo PET imaging methods, such as PiB imaging (Pittsburg Compound B), also provide strong levels of sensitivity and specificity for the detection of AD pathology (Dubois et al., 2007). While imaging techniques can detect and monitor the progression of AD, low-tech tests have also shown to be diagnostically beneficial. For instance, results from the Clock-drawing paradigm, using revised scoring criteria, can differentiate between individuals with Mild Cognitive Impairment who progress to AD from those who do not (Babins et al., 2008). Others have shown that a utilizing a collection of neurospychological tests provides a strong metric for the progression of AD (Behl et al., 2005). However, there is no single test, high-tech or low-tech, that is absolutely favoured for either diagnosing or monitoring AD and thus

2 there is value in investigating new tasks, especially a task that has minimal demands on verbal and manual outputs, such as the antisaccade task. In the antisaccade task, a saccade must be made in the opposite direction of a peripheral target, in contrast with the prosaccade task in which a saccade is made towards a peripheral target (Hallet 1978). Converging bodies of evidence from functional neuroimaging, focal lesion studies and animal models have implicated the dorsolateral prefrontal cortex (DLPFC) as a critical substrate for the successful completion of an antisaccade. During the earliest stages, prior to clinical diagnosis, AD pathology is located in the medial temporal lobes. By the time a clinical diagnosis is made, 3-4 decades of gradually progressing pathology, including extensive neurofibrillary changes (NFC) and amyloid pathology (AP), are found in not only the ventral portions of the brain, but also parietal, temporal and frontal brain regions (Braak & Braak, 1991). The site of the earliest AD pathology, the medial temporal lobes, is consonant with the first diagnostic indicator of AD: progressive losses in episodic memory (Dubois et al., 2007). As the pathology progresses and encompasses additional cortical areas, greater cognitive decline is reported. The DLPFC, critical for antisaccades, is thought to be spared until the mid to final AP and NFC stages. The antisaccade task might provide a functional index of DLPFC impairment, and ultimately DLPFC pathology within AD. For instance, during the earliest stages, minimal or absent DLPFC pathology would be associated with normal antisaccade performance, while in the later stages greater degrees of DLPFC pathology would be associated with increasing numbers of antisaccade errors. Performance on the Mini-mental Status Exam

3 (MMSE) and the AD Assessment Scale Cognitive Subscale (ADAS-cog), which is a widely used psychometric measures of dementia severity, has been shown to correlate with antisaccade error rates: higher levels of dementia were associated with more antisaccade errors (Abel et al., 2002; Boxer et al., 2006; Currie et al., 1991; ShafiqAntonacci et al., 2003). However, past studies have included patients in the more severe stages of AD, resulting in a paucity of data on antisaccade performance in patients with mild AD. Therefore, a primary goal of this study was to examine antisaccade performance in patients with mild AD. A secondary goal was the assessment of within block performance on both the prosaccade and antisaccade tasks in elderly controls and AD patients. Specifically, we examined the difference between trials preceded by a trial of the same direction or of a different direction. Statistical tests utilized in antisaccade studies typically assume that each trial is independent of all other trials. However, there is evidence to suggest that this is not the case, as a trial preceded by a trial of the same direction is associated with fewer errors than a trial preceded by a trial of a different direction. Currently, it is unknown if this effect is present in elderly controls or patients with AD.

4

1.2 Objectives 1.2.1 Determine if mild AD patients make more antisaccade errors than agematched controls Antisaccade performance in AD has been defined by groups of patients with mean MMSE scores between 17 and 20. Although those groups included patients with mild AD, they also included patients with moderate and severe levels of AD, potentially exaggerating the number of errors made by patients with AD. This is because higher degrees of dementia, measured by the MMSE or ADAS, correlated with higher levels of error rates (Crawford et al., 2005; Boxer et al., 2006 ). Thus, the reported error rates in previous AD studies may have been heavily influenced by moderate and severely demented AD patients. Therefore, the first objective of this thesis was to determine if patients with MMSE scores
17 make more errors than age-matched controls using both a laptop and a clinically-based antisaccade task.

1.2.2 Determine if more errors are made on trials preceded by a trial of a different direction relative to trials preceded by a trial of the same direction In young controls, the effect of direction switching from one trial to another causes more antisaccade errors than for direction repetition (see Figure 2) (Reuter et al., 2006; Tatler et al., 2007). However, it is unknown if this effect is found in elderly controls or patients with mild AD, and if this effect is different between the prosaccade and antisaccade tasks. Therefore, the second objective was to determine if more errors are made on different trials relative to same trials, and if this effect is different between elderly controls and patients with mild AD.

5

Figure 1 – Same vs. Different Trials

6

Chapter 2: Literature Review 2.1 Alzheimer’s Disease Alzheimer's disease (AD), characterized by gradual, progressive loss of episodic memory, is the most common single cause of dementia affecting four million Americans, and is quickly becoming one of the "most burdensome health conditions worldwide" (Ferri et al., 2005). In the next two decades, the number of individuals diagnosed with AD will nearly double in North America and Europe, while in Asia it will nearly quadruple (Ferri et al., 2005). As new pharmaceuticals are developed to treat and possibly prevent AD, early diagnosis and disease treatment monitoring will become increasingly important. Currently used NINCDS-ADRDA diagnostic criteria, and the newly proposed criteria (Dubois et al., 2007) both include deficits in episodic memory as the core diagnostic feature of AD. Although decline in episodic memory is central to typical AD, an understanding of additional deficits associated with AD may aid in tracking the progression of AD and monitoring the effectiveness of treatments. Once such deficit that has been noted in patients with AD is having difficulty exerting flexible control over prepotent saccades during the antisaccade task (Table 1). Results from antisaccade studies indicate that the task may have potential for monitoring progression, specifically the emergence of dorsofrontal functional deficits in AD, as well as monitoring new treatments

7

Table 1 - Studies investigating antisaccade performance in Alzheimer’s Disease

N

Control Age

N

Patient Age

Diagnostic Criteria

MMSE

PL

PA

AL

AE

Fletcher et al. (1986)

11

69 (6.5)

13

69 (6.1)

"Clinical Diagnosis"

18.3 (4.1)

ND*

Hypo

NA

74% GT

Currie et al. (1991)

180

41 (18)

30

67 (8)

NINCDS-ADRDA

NA

NA

NA

NA

> 30% GT

Abel et al. (2002)

11

67.4 (5.4)

11

73.1 (9.4)

NINCDS-ADRDA

20.6 (7.6)

ND

NA

NA

75.6% GT

55.9% (32.7)

Shafiq-Antonacci et al. (2003)

245

62.8 (8.6)

35

70.9 (9.4)

NINCDS-ADRDA

17.1 (7.4)

GT

Hypo

GT

ND

53.4% (23.6) Crawford et al. (2005)

18

75.2 (3.8)

18

77.8 (4.8)

NINCDS + DSM IV

20.9 (4.3)

ND

ND

ND

GT

Mosimann et al. (2005)

24

75.3 (5.8)

22

78.1 (6.8)

NINCDS-ADRDA

17.9 (4.7)

ND

ND

ND

80% (42) GT

Boxer et al. (2006)

20

64.4 (7.2)

18

58.4 (7.2)

NINCDS-ADRDA

18.7 (8.5)

ND

ND

ND

~60% GT

Garbutt et al. (2008)

27

65 (1.5)

28

59.8 (1.4)

NINCDS-ADRDA

19.5 (5.3)

GT

ND

GT

~75% GT

MMSE = Minimental Status Exam, PL = Prosaccade Latency, PA = Prosaccade Amplitude, AL = Antisaccade Latency, few studies included antisaccade amplitude, thus it was omitted from the present table, AE = Antisaccade Errors, ND = No difference, GT = Greater than, Hypo = Hypometric, NA = not applicable

8

2.2 The Antisaccade Task In the antisaccade task, a participant is told to inhibit a reflexive, visually-guided saccade to a peripheral target, and to make an antisaccade in the opposite direction to a non-existent target (Figure 2) (Hallett, 1978). Thus, the antisaccade task probes one’s ability to exert flexible control by overcoming the prepotent reflexive saccade response. If the participant fails to inhibit a reflexive saccade and makes a saccade towards the peripheral target, this constitutes an antisaccade (ie inhibition) error. The task has been widely adopted because of several advantages over other tests: it does not require a verbal or manual response and is well tolerated in patients with dementia, including AD, and Frontotemporal Degeneration (FTD) (Mosimann et al., 2005). Furthermore, patients are often unaware of their mistakes and rarely, if ever, become frustrated. Although the task relies on making a parsimonious response - a saccade, multiple, easily quantifiable metrics, such as corrected versus uncorrected errors, saccade amplitude and latency, can be derived from the task. Recently, the neural correlates of the task have become better understood (Munoz & Everling, 2004). An absence of verbal or manual responses enables the antisaccade task to be used in neuroimaging environments, such as magnetoencephelography (MEG) and functional magnetic resonance imaging (fMRI), which do not tolerate movement well and consequently enable non-human primates to complete the task, providing a model for understanding the neural correlates of antisaccades (Munoz & Everling, 2004). The relative simplicity of the antisaccade task has enabled children, adolescents, adults and the elderly to complete the task, which has also provided developmental data.

9

Figure 2 – Prosaccade And Antisaccade Tasks

The arrows denotes where participant should look. During the Prosaccade task they fixate on the cross, then make a saccade to the peripheral target. During the Antisaccade task they fixate on the cross but then look away from the peripheral target.

10

2.3 Developmental Changes The frontal lobes undergo rapid changes from childhood to adolescence, followed by gradual changes during later adulthood as the later evolved structures such as the frontal lobes gradually become fully myelinated. Children under 10 years of age have great difficulty performing the antisaccade task, making more errors in direction than adults (Klein, Foerster, Hartnegg, & Fischer, 2005; Munoz, Broughton, Goldring, & Armstrong, 1998). Between the ages of 10 and 15 antisaccade performance improves dramatically and continues to improve into early adulthood. The reduction in antisaccade error rates appears to closely mirror structural changes that occur from childhood to adulthood. For example, volumetric imaging has shown that between the ages of 8 and 22, total white matter volume increases linearly with age (Giedd et al., 1999). In the frontal lobes, grey matter shows a non-linear increase over time, with a peak volume around the age of 12, followed by a gradual decline presumed to relate to synaptic extermination (Giedd et al., 1999). In addition to volumetric indices of structure, diffusion tensor imaging (DTI) has revealed that fractional anisotropy, a measure of white matter microstructural integrity and myelination, increases in the frontal lobes from childhood to adulthood (Barnea-Goraly et al., 2005). During older adulthood, aging is associated with a general decline in grey matter volume (Good et al., 2001) and a reduction in white matter fractional anisotropy (Pfefferbaum & Sullivan, 2003): a gradual reversal of developmental changes. Studies of antisaccade performance from young adulthood onwards have reported either a nonsignificant upward trend in error rates (Crawford et al., 2005; Currie, Ramsden, McArthur, & Maruff, 1991; Munoz et al., 1998; Raemaekers, Vink, van den Heuvel,

11 Kahn, & Ramsey, 2006), or a significant increase in error rates with aging (Abel & Douglas, 2007; Klein et al., 2005; Olincy, Ross, Youngd, & Freedman, 1997). Furthermore, fMRI has revealed a compensatory shift in antisaccade related activation between young adults and older adults indicative of functional differences between young and older adults (Raemaekers et al., 2006).

2.4 Functional Imaging Studies Developmental and aging studies indicate a relationship between frontal lobe function and inhibitory failure in the antisaccade task. Functional neuroimaging techniques such as positron emission tomography (PET) and fMRI have provided more specific information on neural substrates, implicating specific regions within the frontal lobes for successful antisaccade generation. In the simplest imaging experiments using either PET or fMRI, blocks of antisaccades were compared to prosaccades, and revealed greater activation for antisaccades in the frontal eye fields and the superior parietal lobule, when compared with prosaccades (Kimmig et al., 2001; O'Driscoll et al., 1995). Early functional imaging studies led to conflicting views on the involvement of the dorsolateral prefrontal cortex (DLPFC): some found activation in the right DLPFC (DeSouza, Menon, & Everling, 2003; McDowell & Clementz, 2001), while others did not (Kimmig et al., 2001; O'Driscoll et al., 1995). However, block designs have many shortcomings for analysing antisaccade related activity. Block designs do not allow temporal differentiation between the components of an antisaccade such as: 1) inhibiting a prosaccade, 2) generating an antisaccade and 3) making a retrosaccade back to the central fixation. There is the potential that activation could either be nullified or enhanced

12 by a negative or a positive response respectively, from one of the other components. Furthermore it is difficult to tease apart the effect of antisaccade directional errors, reduced latencies and hypometric responses on the blood oxygen level dependent (BOLD) signal. Event-related designs, in which each antisaccade event is temporally spaced, and later averaged, have been used to overcome these problems. Using an event-related design, the BOLD response associated with either preparation or the motor phase of an antisaccade can be compared. When the BOLD response during these two phases was examined, it was discovered that increases in BOLD signal in the right DLPFC and bilateral frontal eye fields, associated with antisaccades, were actually due to preparation and not action (DeSouza et al., 2003). These findings were confirmed in a mixed electroencephalogrophy (EEG)/MEG study which also found more activity in the medial aspects of the frontal eye fields, supplementary eye fields and prefrontal cortex during the planning phase of an antisaccade (McDowell et al., 2005). Finally, during the retrosaccade, when the participant makes a saccade back to the center, there is actually a negative bold response (Raemaekers, Vink, van den Heuvel, Kahn, & Ramsey, 2005), suggesting a mechanism by which retrosaccade activity may have cancelled out positive DLPFC BOLD signal in some block designs. Comparison between successful antisaccades and antisaccade errors has also revealed some important details. For instance, fMRI and EEG have revealed increased BOLD signal and increased negative potentials, respectively, in the DLPFC during correct antisaccades when compared with incorrect (Everling, Spantekow, Krappmann, & Flohr, 1998; Ford, Goltz, Brown, & Everling, 2005). When Ford and colleagues (2005) compared correct antisaccades with correct prosaccades, they found greater activity in the

13 anterior cingulate cortex, left DLPFC, bilateral frontal eye fields, pre-supplementary eye fields and parietal areas. When they compared correct to incorrect antisaccades, they found a greater BOLD response in the right DLPFC, anterior cingulate cortex and presupplementary eye fields. Thus, the left DLPFC shows greater activity for antisaccades compared with prosaccades, and the right DLPFC exhibits greater activity for correct than incorrect antisaccades. The differences between right and left DLPFC activation reported by Ford and colleagues (2005) parallel the localization of two of the three frontal processes described by Stuss and Alexander (2007), task setting and task monitoring. The antisaccade task requires greater task-set demands than the prosaccade task, which may explain the greater left DLPFC activation for antisaccades relative to prosaccades. In contrast to task setting, involving the left DLPFC, Stuss and Alexander (2007) reported an association between right lateral regions and impairments in task monitoring. Greater right DLPFC activation for correct antisaccades, relative to incorrect antisaccades, may reflect increases in task monitoring related to successful antisacccade performance.

2.5 Focal Lesion Studies The involvement of the DLPFC in antisaccades can inferred indirectly from functional neuroimaging studies, but data from patients with focal lesions provide more direct evidence that its integrity is necessary for making correct antisaccades. Lesions affecting either the left or right mid-DLPFC are consistently associated with increased error rates (Guitton, Buchtel & Douglas, 1985; Pierrot-Deseilligny, Rivaud, Gaymard, & Agid, 1991; Ploner, Gaymard, Rivaud-Pechoux, & Pierrot-Deseilligny, 2005). For example, Pierrot-Deseilligny and colleagues (1991) examined patients with lesions that

14 affected either the posterior parietal cortex, the frontal eye fields, supplementary motor area or the DLPFC. They found that patients with lesions in the right or left DLPFC made more directional errors than patients with lesions in the frontal eye fields or supplementary motor area, and than controls. In a recent study, patients with lesions resulting from infarcts were separated into two groups: those whose antisaccade performance was in the same range as normal controls and those who fell out of that range (Ploner et al., 2005). Lesion analysis showed that damage in the high error group primarily involved areas in either the right or left dorsofrontal cortex that have efferent connections through the anterior limb of the internal capsule with the superior colliculi. Specifically the mid-DLPFC, which has efferent connections through the anterior limb of the anterior capsule into the superior colliculi, was the only area that was damaged in each patient in the high error group. Although Ploner and colleagues (2005) did not conduct significance tests on latency differences between the two groups, the high error group showed longer latencies than the low error group. Amplitude was not reported. Gaymard and colleagues (Gaymard, Francois, Ploner, Condy, & Rivaud-Pechoux, 2003) observed that a patient with a small focal lesion affecting the connections between the DLPFC and superior colliculi made more directional errors than controls, but had normal latency and amplitude. A subsequent study that examined 30 patients with subcortical lesions found that only those with damage to the anterior limb of the internal capsule, the genu or the most anterior portion of the posterior limb of the internal capsule had high error rates. In contrast, those not considered impaired on the task had no damage to those regions; rather, their damage involved the posterior limb of the internal capsule and parts of the thalamus and basal ganglia (Condy, Rivaud-Pechoux, Ostendorf, Ploner,

15 & Gaymard, 2004). Both groups of patients showed similar latencies and amplitude during the prosaccade task, but latency and amplitude were not reported for the antisaccade task. The fact that there is some understanding about the neural correlates of the antisaccade task coupled with its other advantages make the antisaccade task an attractive measure of inhibitory control and executive function. Given that lesion and functional imaging evidence both support a critical role of the DLPFC in the antisaccade task, the task may also provide an ideal index of DLPFC integrity in the dementias, such as AD.

2.6 Differentiating Alzheimer’s Disease From Healthy Aging Patients with AD make significantly more antisaccade errors than controls (Table 1) and also leave many more of their errors uncorrected (Boxer et al., 2006). During the prosaccade task (Figure 1), patients with AD perform mostly within normal range (Table 1), emphasizing that errors in the antisaccade task are made in the absence of any impairment in visually-guided saccades. Despite the fact that AD patients make significantly more antisacccade errors than age-matched controls, antisaccades have limited diagnostic potential in differentiating patients from age-matched controls. ShafiqAntonacci and colleagues (Shafiq-Antonacci, Maruff, Masters, & Currie, 2003) found that although robust differences between patients with AD and controls exist, the antisaccade task had only a modest sensitivity and specificity and concluded that "antisaccade performance cannot identify AD in individual cases". In support of this conclusion, results from the most optimistic studies indicate that the task can only

16 differentiate 40% of patients with AD from age-matched controls (Table 2). Moreover, groups of patients with AD show a much larger variance in the percentage of antisaccade errors relative to age-matched controls (Abel, Unverzagt, & Yee, 2002; Fletcher & Sharpe, 1986; Shafiq-Antonacci et al., 2003), emphasizing that some patients are either not significantly impaired or are not impaired at all. Relative to tests of episodic memory, antisaccades offer little utility in the detection of AD, however they may be an index of DLPFC involvement and they may be useful for monitoring emergence of DLPFC deficits, monitoring progression and possibly response to treatment in AD.

Table 2 - Power Analysis and Differentiation Capabilities of the Antisaccade task in Alzheimer’s Disease

Control Errors (SD) AD Errors (SD) Mean Difference Cohen’s d*

Effect Size**

Shafiq-Antonacci

31.44 (15.36)

55.88 (32.74)

24.44

0.96

0.43

Crawford

18.4 (13.4)

53.4 (23.6)

35

1.83

0.67

Mosimann

25 (38)

80 (42)

55

1.37

0.57

Only studies which provided both mean and standard deviation (SD) values were included in this table *Cohen’s d = (MeanAD – MeanCon)/SQRT((SD2AD + SD2CON)/2) **Effect size = (Cohen’s d)/SQRT((Cohen’s d)2 + 4)

17

2.7 Antisaccade Errors and Dorsolateral Prefrontal Pathology in AD Declines in hippocampal volume and episodic memory appear to be the earliest brain-behaviour correlations in AD, mirroring the development of AD pathology (Dubois et al., 2007). Progression and topographical distribution of neuropathology in AD has been categorized into six stages, in which AD neurofibrillary tangles, first appears in the limbic system beginning in the medial temporal lobes and then gradually moves into temporal-parietal association cortices and then the frontal regions (Braak & Braak, 1991). The DLPFC, critical for antisaccades, is thought to be relatively free of tangle pathology during the early phases of the disease, though paradoxically, amyloid PET labelling suggests presence of amyloid even in preclinical stages of the disease (Pike et al., 2007). Therefore patients with early AD, who may have little or no dorsal frontal pathology, may show little to no impairment on the antisaccade task. The mini-mental status exam (MMSE), a general measure of cognition, shows a strong correlation with antisaccade error rates: as MMSE scores decline and dementia worsens, patients make more errors (Abel et al., 2002; Boxer et al., 2006; Currie et al., 1991; Shafiq-Antonacci et al., 2003). Boxer and colleagues (2006), examined a subgroup of patients with mild AD, who had MMSE scores greater than 22 out of 30, and found that they did not make significantly more errors than controls, suggesting antisaccade impairments may not arise in the early stages of the disease. An absence of antisaccade impairments in patients with mild AD, followed by a gradual increase in errors with dementia severity, as indexed by the MMSE, suggests that the task may provide information on not only the progression of the disease but also the progression and magnitude of DLPFC functional impairment.

18

2.8 Barriers in Adopting the Antisaccade Task as a Probe of DLPFC Function The validity of using the antisaccade task as a measure of executive function and inhibitory control in patients with AD is supported by its correlation with other measures of executive function such as the Stroop task and the Trials A and B tasks (Boxer et al., 2006; Crawford et al., 2005). However, before the antisaccade task can be used as a clinical index of DLPFC function in AD, there are several issues that require further clarification. First, the relationship between dementia severity and antisaccade performance is based on cross-sectional between-subject data, not longitudinal withinsubject data. Second, the proposed relationship between DLPFC pathology and increased error rates in AD is largely based on indirect inference and would benefit further from confirmatory structural and functional neuroimaging. Third, a successful antisaccade consists of several underlying processes, such as inhibition of a prepotent saccade and generation of a voluntary saccade, and impairment in any of the subprocesses may result in increased error rates, obfuscating potential brain behaviour relationships. Furthermore, despite the potential testing convenience provided by clinical versions of the antisaccade task, which will be discussed shortly, typical saccade measurement relies on costly eyemonitoring equipment not available in most clinical environments, potentially deterring widespread clinical adoption.

19

2.9 The Antisaccade Task as an Index of Severity The relationship between MMSE scores and antisaccade error rates supports the notion that error rates may mirror disease progression, but are 1) based on cross-sectional comparisons, and 2) under-represent those in the earliest stages of AD. Most existing studies have focused on patients in the moderate to severe stages of AD, which may exaggerate differences between AD and controls and may strengthen the correlation between MMSE scores and error rates. For instance patients on the lowest end of the MMSE spectrum tend to make 100% uncorrected errors making it difficult to infer whether directions were even understood or remembered. An emphasis on mild to moderate patients may reduce the likelihood of including patients who fail to understand the task and could enhance the understanding of the relationship between antisaccade performance and severity of dementia. Secondly, patients in the earlier stages of AD show greater heterogeneity in neuropsychological impairment. For instance, some AD patients have been categorized as “frontal variant AD” as they not only perform significantly worse on frontal tasks, than average AD patients, but also show greater frontal atrophy (Johnson, Head, Kim, Starr & Cotman, 1999). Inclusion of a small number of mild AD patients may have failed to capture variations in frontal deficits potentially underestimating error rates in mild AD.

2.10 Neuroimaging & Dementia The relationship between AD neurofibrillary tangle pathology distribution and antisaccade deficits is largely based on an assumption that those who are less demented,

20 and make fewer errors, may have less pathology in the DLPFC. Boxer and colleagues (2006) used voxel-based morphometry to examine the association between brain atrophy and antisaccade performance in patients with AD and those with frontotemporal degeneration (FTD). They found that the volume of an area located ventrally to the right frontal eye fields was correlated with correct responses, while supplementary eye fields volume was correlated with antisaccade latency. Although these findings reinforce the relationship between the dorsofrontal cortex and antisaccade performance, they failed to find a relationship between antisaccade performance and DLPFC or frontal eye field volumes, which as, stated earlier, are regions critical for inhibition of prosaccades and generation of antisaccades respectively. By including both FTD and AD patients in a single group, despite the heterogeneity in the distribution and type of pathology within FTD and AD (Josephs et al., 2006; Snowden, Neary, & Mann, 2007), Boxer and colleagues (2006) may have prevented detection of correlations between structure and performance. For instance, high variability in antisaccade performance, coupled with the low sensitivity of antisaccades (Shafiq-Antonacci et al., 2003) indicates many patients are unimpaired and may lack sufficient DLPFC pathology and atrophy, thus eliminating any significant correlations between structure and function. Furthermore, Boxer and colleagues (2006) corrected for MMSE scores, which are strongly correlated with error rates in AD; controlling for MMSE scores may have inadvertently eliminated any relationship between error rates and DLPFC or frontal eye fields volumes. Utilizing methodology from antisaccade focal lesion studies (Ploner et al., 2005), future investigations should divide patients with AD into two groups: 1) those whose percentage of errors fall within normal range and 2) those who make significantly more errors than

21 normal. This design would strengthen brain behaviour correlations by reducing the heterogeneity of performance and presumably pathology distribution within AD groups. Diffusion tensor imaging, a technique for examining white matter tissue microstructure, has revealed differences in patients with AD relative to controls. Specifically, the superior longitudinal fasciculus, a bundle of fibers that connects posterior and frontal regions of the brain and also includes connections to the DLPFC, was reported to be associated with decreases in the Fractional Anisotropy (FA) (Medina et al., 2006). A decrease in FA, a measure of white matter cohesion and integrity, is indicative of underlying structural aberrations that may elucidate brain-behaviour correlations. Correlating error rates and diffusion tensor metrics such as FA would provide insight into the potential role white matter injury may have on antisaccade performance in AD. Functional imaging techniques such as single photon emission computerized tomography (SPECT), a semiquantitative measure of regional cerebral blood flow and fMRI have provided important insights into the relationship between memory and brain function in AD and may provide further insight into antisaccade errors. Using SPECT, Garrido and colleagues (Garrido et al., 2002) found that patients with AD showed decreased cerebral blood flow in the left medial temporal lobes, relative to controls during a verbal recognition memory task. Likewise Grady and colleagues (Grady et al., 2003) used fMRI to study compensatory frontal network activity during mental tasks in mild AD. The relationship found between functional aberrations and poor memory might be mirrored in a relationship between increased antisaccade errors and abnormal DLPFC activation patterns. For instance, fMRI and positron emission tomography have revealed

22 decreases in activation in the anterior cingulate and DLPFC in Schizophrenia, a psychopathology associated with increased antisaccade errors (Hutton & Ettinger, 2006). In AD, functional imaging results coupled with structural neuroimaging could provide more direct in vivo evidence for a correlation between DLPFC changes and increased error rates.

2.11 Fractionation of Processes in the Antisaccade Task The antisaccade task is comprised of multiple sub-processes that contribute to a successful execution (Munoz & Everling, 2004). A participant must be able to fixate on the central fixation point, then inhibit a reflexive saccade, invert the saccade vector to a non-existent target and make a voluntary saccade in the direction of the updated vector. In addition to the processes directly related to the task, there are secondary processes that are also critical for successful antisaccades. For instance, a patient must be able to understand and remember the task's directions and vigilantly attend to the task. According to the accumulator model, the process that reaches threshold first, either the antisaccade or the erroneous prosaccade, determines which behaviour is initiated (Munoz & Everling, 2004). For instance, if the antisaccade process reaches threshold before an erroneous prosaccade, an antisaccade is carried out. Deficits in any of the abovementioned processes could result in a slowing of the antisaccade process, increasing the chance that an erroneous prosaccade is generated. In determining whether antisaccade errors in AD result from inhibitory deficits and ultimately DLPFC dysfunction, it will be critical to fractionate the underlying sub-processes to insure other components impaired in AD, such as memory, are not contributing significantly to antisaccade errors.

23

2.11.1 Fractioning the Antisaccade Task: Inhibition Control Fractionation of inhibition from the other processes can be accomplished through additional saccade task manipulations. Two tasks which focus on inhibition are: 1) the no-go task, in which the participant maintains fixation while peripheral targets appear, and 2) the go no-go task, in which the participant must fixate during some trials (no-go trials) and make saccades during other trials (go trials). No-go and go no-go tasks are easier than the antisaccade tasks because they do not require a vector inversion or volitional generation of a saccade: they simply require inhibition in a proportion of the trials. Crawford and co-workers (Crawford et al., 2005) found that patients with AD made significantly more inhibition errors on both tasks, when compared with controls, suggesting that inhibitory deficits, not deficits in vector inversion or volitional control, may be the greatest contributor to antisaccade errors in AD.

2.11.2 Fractioning the Antisaccade Task: Fixation Typical antisaccade coding schemes require a participant to fixate centrally for at least 200 ms prior to a successful antisaccade; trials that fail to meet this requirement are eliminated from analysis (Shafiq-Antonacci et al., 2003). Reported fixation deficits in patients with AD (Fletcher & Sharpe, 1986) may result in greater numbers of eliminated trials in the AD group, resulting in fewer analyzable trials. Presenting a static image with several targets and asking the participant to maintain fixation on a single target for a

24 specified time provides a parsimonious method for testing fixation. The task could be repeated using different static images and different fixation points.

2.11.3 Fractioning the Antisaccade Task: Vector Inversion Successful antisaccades require a vector inversion: the prosaccade vector must be inverted to a saccade in the opposite direction towards a non-existent target, a visuospatial process which imaging and lesion studies have suggested is mediated by the posterior parietal cortex (Medendorp, Goltz, & Vilis, 2005; Nyffeler, Rivaud-Pechoux, Pierrot-Deseilligny, Diallo, & Gaymard, 2007). In the early stages of the clinical presentation of AD, neurofibrillary tangle pathology is found in the posterior parietal lobe, suggesting a possible association between visuospatial/vector inversion deficits and brain pathology. Although not typically the presenting symptom, spatial and visuomotor deficits are often detectable in the mild to moderate stages of AD (Buck, Black, Behrmann, Caldwell, & Bronskill, 1997; Cummings & Cole, 2002; Tippett & Sergio, 2006). Correlations of performance of antisaccade tasks have been reported with two visuospatial tasks: figure copying with correct antisaccades (Boxer et al., 2006) and spatial span with uncorrected errors (Crawford et al., 2005). Despite these correlations of visuospatial deficits with antisaccade errors, it is unclear if they are contributing factors to the increased error rates present in AD. For instance, error rates are significantly elevated during the no-go task, which only includes the inhibitory component of the antisaccade task, not the visuospatial component. Identifying whether visual spatial deficits contribute to increased error rates in AD require further investigation.

25

2.11.4 Fractioning the Antisaccade Task: Voluntary Saccades A voluntary saccade to a non-existent target differs both neurologically and in difficulty from the reflexive saccades generated in the prosaccade task. While the frontal eye fields are involved in generating voluntary saccades, a region in the parietal lobes is implicated in generating reflexive visually guided saccades (Munoz & Everling, 2004). When patients with frontal eye field focal lesions make antisaccades, they typically show longer antisaccade latency while error rates remain normal (Gaymard, Ploner, RivaudPechoux, & Pierrot-Deseilligny, 1999). In contrast, patients with AD do not consistently show longer latencies, but do consistently make more errors. Furthermore, the frontal eye fields are one of the last regions to be affected by AD tangle pathology making it further unlikely that voluntary saccade generation is impaired or is the cause of increased error rates in AD.

2.12 Memory, Understanding & Attention A failure to make correct antisaccades could result from deficits in memory, understanding and attention. Despite significant impairments in each of those domains in AD, studies have posited that those impairments do not contribute to increased errors because of two factors. First, when performance on an antisaccade block is divided into two halves, performance is stable between the two halves; indicating patients are not progressively forgetting task instructions (Crawford et al., 2005). Second, patients usually generate at least one correct antisaccade, or one corrected antisaccade, a reaction not seen

26 in prosaccades. This suggests that patients remembered the task well enough to make at least one antisaccade (Fletcher & Sharpe, 1986). However, just as inhibitory control is considered impaired in patients with DLPFC lesions, despite generating a few correct responses, patients with AD may still make errors due to poor short-term memory. Neuropsychological indices of memory, such as verbal episodic memory (Boxer et al., 2006) and memory quotient (Currie et al., 1991), correlate with antisaccade performance, suggesting that impairments in memory may partly contribute to antisaccade errors. Perhaps the most pertinent question is whether patients not only remember and understand the task’s instructions immediately before the task begins, but also once it has ended. A patient’s understanding of the task is often determined by having the patient either point to where they are suppose to look, or verbal repetition of the instructions (Crawford et al., 2005). Repeating these steps after the task is completed would not only test memory but would also test a patient’s understanding of the task, but has yet to be accomplished. Although deficits in attention have been documented in patients with AD (Perry & Hodges, 1999), their effect on antisaccade performance has not been explored. If impairments in memory, understanding or attention lead to increased errors, there would be little utility in using the antisaccade task as a measure of disease progression as memory tasks would suffice. If either impairments in memory, understanding or attention contribute to increased error rates, a link between the DLPFC, inhibitory control and error rates in AD would be difficult to infer. Thus, excluding these processes and behavioural domains as major contributors to antisaccade error rates should remain a priority for future studies.

27

2.13 Task Sequence Statistical analysis of antisaccade data has primarily relied on univariate models that assume performance on a single trial is independent from other trials. For example, these models assume that performance on a trial (n + 1) is in no way affected by the previous trial (n). However, this assumption appears false as data from normal controls indicates performance on a current trial is influenced by the direction of the previous trial, relative to the direction of the current trial (Barton et al., 2006; Reuter et al., 2006; Tatler et al., 2007). For instance, if a peripheral target appears on the left side for two trials in a row the second trial would be 1) categorized as “same” and 2) would be associated with fewer errors than a “different” trial (a trial preceding by a trial of a different direction, i.e. left then right) (Barton, Goff & Manoach, 2006). However, this effect may not be consistent across age and patient groups, potentially confounding differences between patients with AD and elderly controls.

2.14 Clinical Adaptation Although clinical variations of the antisaccade task are easy to administer, typical antisaccade experiments use sophisticated eye-tracking labs that are often costly to establish, lack portability and use techniques that require calibration, making the task clinically less appealing. In an effort to avoid these shortcomings, Currie and colleagues (1991) developed a clinical version of the antisaccade task that uses the clinician's nose as the central fixation and fingers as the peripheral targets. Although the clinical variant

28 yielded slightly lower error rates than the laboratory version, scores were highly correlated with those generated in the laboratory version (r = 0.921). The clinical version is also a component of the HIV Dementia scale that has been validated in both patients with HIV (Power, Selnes, Grim, & McArthur, 1995) and patients with subcortical vascular cognitive impairment (van Harten, Courant, Scheltens, & Weinstein, 2004). The clinical variant of the antisaccade task provides a parsimonious method for testing and overcomes the shortcomings of the laboratory version. However, it is unclear how difficult it is for a clinician to remember a sequence of 10 to 20 trials while keeping track of how many trials have been administered and how many errors a patient has made. The clinical version, in its present form, limits the clinician to record only errors in direction, neglecting other metrics such as uncorrected errors, fixation errors, errors of omission, latency and amplitude partly because of the cognitive load of administering the task, but also because latency and amplitude can only be reliably recording with eye-tracking equipment. Although the reliability of the clinical version, between different clinical centers, has not been tested, the laboratory version has been tested in patients with schizophrenia, revealing that the task can be carried out reliably in different hospitals (Radant et al., 2007).

2.15 Non-Alzheimer’s Dementia The majority of studies examining antisaccade performance in patients with dementia have focused on Alzheimer’s disease. However, antisaccade performance has been examined in other types of dementia (Table 3). A pattern of inhibitory impairment, as measured by antisaccade errors, seems to mirror the distribution of pathology in

29 frontotemporal dementia, semantic dementia and progressive non-fluent aphasia. Frontotemporal dementia and progressive non-fluent aphasia, characterized by deficits in behavioural regulation and non-fluent speech respectively are both associated with pathology in the frontal lobes (Neary et al., 1998). In contrast, the core diagnostic feature of semantic dementia is a deficit in word comprehension and it is associated with pathology in the anterior temporal lobes. The difference in the presence of frontal pathology between these groups is mirrored by their performance on the antisaccade task: frontotemporal dementia and progressive aphasia are associated with high antisaccade error rates, while semantic dementia is not (Boxer et al., 2006). Interestingly, all three of these groups correct significantly more errors than patients with AD (Garbutt et al. 2008).

30 Table 3 - Studies investigating antisaccade performance in non-Alzheimer’s Dementia

N

Control Age

N

Patient Age

Diagnostic Criteria

MMSE

PL

PA

AL

AE

Meyniel et al. (2005)

10

68 (9)

23

67 (9)

Lund-Manchester

26.1 (2.8)

GT

NA

GT

63% GT

Boxer et al. (2006)

20

64.4 (7.2)

14

59.9 (5.6)

Neary et al. (1998)

25.7 (3.7)

ND

ND

ND

60% GT

Garbutt et al. (2008)

27

65 (1.5)

24

57.4 (1.7)

Neary et al. (1998)

23.5 (7.5)

NA

Hypo

ND

~55 % GT

Boxer et al. (2006)

20

64.4 (7.2)

7

65.7 (7.9)

Neary et al. (1998)

23.1 (5.3)

ND

ND

ND

~70% GT

Garbutt et al. (2008)

27

65 (1.5)

6

64.5 (3.0)

Neary et al. (1998)

25.2 (3.5)

ND

ND

ND

~52% GT

Boxer et al. (2006)

20

64.4 (7.2)

10

60.3

Neary et al. (1998)

20.1 (7.6)

ND

ND

ND

~25% ND

Garbutt et al. (2008)

27

65 (1.5)

19

60.3 (1.3)

Neary et al. (1998)

21.7 (7.3)

ND

ND

ND

~28% ND

Vidaihet et al. (1994)

12

63.9 (8.3)

10

62.5 (5.5)

NA

NA

ND

Hypo

NA

~ 73.5 GT

Meyniel et al. (2005)

10

68 (9)

14

70 (6)

Litvan's

28.6 (2.0)

ND

NA

GT

70% GT

Rivaud-Pechoux (2007)

10

64 (9)

12

66 (9.9)

Litvan's

NA

ND

NA

GT

GT

Garbutt et al. (2008)

27

65 (1.5)

10

65.5 (1.3)

Litvan et al. (1996)

26.8 (2.6)

GT

NA

NA

~92% GT

Vidaihet et al. (1994)

12

63.9 (8.3)

10

66.5 (6.8)

NA

NA

GT

ND

NA

~ 37.5 ND

Rivaud-Pechoux (2007)

10

64 (9)

8

76 (5.4)

Litvan's 1997

NA

GT

NA

GT

ND

Garbutt et al. (2008)

27

65 (1.5)

15

62.7 (2.0)

Several methods

19.8 (7.7)

GT

Hypo

NA

~75% GT

FTD

PNFA

SD

PSP

CBD

MMSE = Minimental Status Exam, PL = Prosaccade Latency, PA = Prosaccade Amplitude, AL = Antisaccade Latency, few studies included antisaccade amplitude, thus it was omitted from the present table, AE = Antisaccade Errors, ND = No difference, GT = Greater than, Hypo = Hypometric, NA = not applicable

31 Although Corticobasal Syndrome and Progressive Supranuclear Palsy are primarily characterized by motor impairments such as asymmetric extrapyramidal signs and vertical gaze palsy respectively, they are also associated with frontal deficits (Kertesz, Martinez-Lage, Davidson, & Munoz, 2000; O'Keeffe et al., 2007). The clinical criteria for diagnosis of Progressive Supranuclear Palsy is highly predictive of autopsy findings, whereas Corticobasal Syndrome is less specific and can be associated with either corticobasal degeneration, progressive supranuclear palsy pathology, both overlapping (Josephs et al., 2006) or even the tau negative ubiquitin positive pathology (Masellis et al., 2006). Patients with Corticobasal Syndrome and controls made an equal number of errors on the antisaccade task, while patients with Progressive Supranuclear Palsy made significantly more errors than controls (Meyniel, Rivaud-Pechoux, Damier, & Gaymard, 2005; Vidailhet et al., 1994). However, when patients with CBD completed a mixed block of prosaccades and antisaccades, they made many more errors than controls compared with patients with PSP who showed no difference between mixed and non-mixed blocks (Rivaud-Pechoux, Vidailhet, Brandel, & Gaymard, 2007). Differences in antisaccade performance between the two groups may provide information to help distinguish these diseases.

2.16 Discussion Pressure to increase the diagnostic accuracy for dementia, and specifically Alzheimer’s disease is mounting due to the availability of new potential treatments. In the early stages, when the drugs will be the most effective, it is unlikely that the antisaccade task will provide greater diagnostic utility than existing tests for AD. However, in

32 addition to its potential role as a probe of dorsolateral function and as a test to monitor treatment response, it may aid in differentiating other forms of dementia.

2.17 Conclusion The neural correlates of antisaccades continue to be mapped and reported using a variety of neuroimaging techniques providing further insights into brain behaviour correlation of this simple task. Antisaccades provide a well tolerated, language-free and hands-free neuropsychological probe that may not only help those with AD, but could be especially helpful in testing patients with expressive language problems or motor deficits, such as those with Progressive Aphasia and Amyotrophic Lateral Sclerosis. Despite having limited utility in differentiating individuals with AD from normal aging the available evidence indicates that the task may provide insight into frontal lobe function and an index of DLPFC pathology in AD. The potential utility of the antisaccade task as a neuropsychological probe of DLPFC and ultimately for progression and for treatment monitoring appears promising but requires further investigation. The relationship between the DLPFC, inhibitory control and errors rates in AD requires exclusion of other potential contributors, such as memory impairments, and would benefit from converging evidence from multiple neuroimaging modalities. This overview has outlined future avenues of research for testing the link between inhibitory deficits and DLPFC changes in patients with AD.

33

Chapter 3 3.1 Background The antisaccade task has been used increasingly to study Alzheimer’s disease (AD) because it provides a parsimonious hands and language free measure of dorsolateral prefrontal cortex (DLPFC) function (Garbutt et al., 2008). In the antisaccade task, an eye movement must be directed in the opposite direction from a rapidly appearing target (Hallet 1978). Healthy individuals typically make antisaccade errors (looking towards the target) on 20% of trials, while patients with AD make between 50 to 80% errors (Crawford et al., 2005; Garbutt et al., 2008). Previous studies have included AD patients ranging from mild to severe levels of dementia with mean Mini Mental Status Exam (MMSE) scores between 17 and 21 (see Table 1). Reports of negative a correlation between MMSE and antisaccade error rates (low MMSE scores corresponds with high error rates)(Currie et al., 1991; Shafiq-Antonacci et al., 2003) suggests that the inclusion of more severely demented patients may have exaggerated the differences in error rates between patients and controls. In addition to differences between patients and controls, there are differences in error rates between trials preceded by trials of the same direction (i.e. two right-directed trials in a row) compared with trials preceded by trials of a different direction (i.e. a leftdirected trial followed by a right-directed trial) (see Figure 2). Specifically, healthy young participants made more antisaccade errors on "different" trials (i.e. left trial followed by a right trial) relative to "same" trials (i.e. left trial followed by a left trial) (Barton et al.,

34 2006). However, it remains unknown if this effect is present in elderly controls or patients with AD. A clinical version of the antisaccade task, in which the clinician’s nose and index fingers become the central fixation point and peripheral targets respectively, may make antisaccade testing more accessible by providing an inexpensive, portable and easy to use method of assessing antisaccade performance that has already been validated in clinical populations (Currie et al., 1991).Therefore the first objective was to assess the clinical version of the antisaccade task in patients with AD and age-matched elderly controls. The second objective was to determine if patients with mild AD (MMSE
17), relative to previous studies that included patients with more severe AD (MMSE
12), a) made more errors than controls and if b) error rates were correlated with neuropsychological measures such as the MMSE or the dementia rating scale (DRS). The third objective was to determine if elderly controls or patients with AD made more errors on different trials relative to same trials.

3.2 Hypotheses 1) A clinical alternative to conventional versions of the antisaccades task will provide a parsimonious and valid measure of antisaccade performance in elderly control and patients with AD. Specifically, antisaccade error rates on the clinical version will correlate strongly with error rates from a laptop-based version of the task. 2)

a) Patients with mild AD (mild relative to previous studies) will make

significantly more errors than age-matched controls

35 b) Antisaccade error rates within the AD group will significantly correlate with dementia severity. Specifically, lower MMSE scores will be related to higher levels of antisaccade errors. 3) Both patients with AD and elderly controls will make more errors on different trials (trial preceded by a trial of a different direction) relative to same trials (trials preceded by a trial of the same direction) during the antisaccade task. However, there will be no differences in error rates between the two trial types during the prosaccade task.

3.3 Methods 3.3.1 Study Participants Thirty Patients with AD and thirty-one age-matched controls were drawn from the Sunnybrook Dementia Study, a large longitudinal clinical and multimodal imaging study of Alzheimer’s and other dementias. Patients were diagnosed with probable AD using the NINCDS-ADRDA criteria and the DSM-IV criteria for dementia. Patients with neurological or psychiatric conditions including major depression were excluded. Patients with MMSE scores less than 17 were also excluded. A cut-off score of 17 was chosen because 16 roughly corresponds to an inflection point in which the slope of cognitive decline increases significantly (Feldman et al., 2001). All patients and controls completed an MMSE, while only a subset of patients in this particular study completed the following neuropsychological measures: DRS, Disability Assessment for Dementia (DAD) scale (Gauthier, Gélinas, & Gauthier 1997), Wisconsin Card Sorting Task (WCST) (Milner, 1963), Trials A & B and the backwards digit span. All participants, or

36 their designated substituted decision maker, provided informed consent prior to participating in the study. Although the proportion of female and male participants was unequal between the two groups it is unlikely that this would affect error rates as there has been no evidence that sex differences exists in antisaccade performance (Ettinger et al., 2005).

3.3.2 Saccade Tasks Laptop Version: The laptop tasks were presented on a laptop with a 15.4 screen running at resolution of 1440 by 900 pixels. Both the black fixation star and the black peripheral target (circle) were roughly 75 pixels by 75 pixels and were presented on a light gray background. The peripheral targets were presented 500 pixels to the left or right of center. Each participant completed one block of prosaccades and two blocks of antisaccades (24 trials per block) (See Figure 3). Fixation lasted for 2000ms, then simultaneous to the disappearance of the fixation star, a peripheral target appeared to the left or right of center for 1000ms. Timing values were chosen as they represented a rough mean of the methodology utilized in the previous 8 studies examining antisaccade performance in AD. The step paradigm, in which the central fixation point disappears in synchrony with the peripheral target’s appearance, was chosen for two reasons. First, it represents a temporal compromise between the gap or overlap paradigms, and second it was presumed that because fewer errors are typically made during the step paradigm, relative to the gap paradigm, elderly controls and patients with AD would be less frustrated and more compliant. The timing and stimulus of both the prosaccade task and the antisaccade task were identical; the tasks only differed in the instructions given to the participant prior to

37 each block. Although the distance between the center and the peripheral target was held constant during each trial, participants were able to move their head freely; thus the visual angle of the offset was not equal for each participant. Prior to each block, participants were asked to look at the fixation point, then in the case of the prosaccade task, look towards the peripheral target, or in the case of the antisaccade task, look away from the target. To demonstrate an understanding of the antisaccade task, prior to the first block, participants first had to successfully point to the location where they were supposed to look for three consecutive trials. A laptop-integrated web camera recorded the participants’ actions. Clinical Version: After being administered the laptop version, participants completed a clinical version of the test, which consisted of one block of prosaccades followed by one block of antisaccades (10 trials per block). Participants were instructed to look at the examiner's nose (central fixation point), then look towards the hand with the raised index finger (for the prosaccade block) or away from the hand with the raised index finger to the opposite hand during the antisaccade block. The same sequence (R, L, L, R, L, R, R, L, R, L) was given to every participant for each block. During the task, the examiner mentally noted the number of errors and recorded them upon completion of a block. Each participant was seated approximately 1 meter from the clinician. Same trials & different trials: The effects of direction, from one trial to the next, on saccade responses were analyzed by categorizing each trial as same or different. A trial was categorized as same if the peripheral target was offset in the same direction as the previous trial. In contrast, a trial was categorized as different if the peripheral target was offset in a different direction as the previous trial (see Figure 2.).

38

Figure 3 – Laptop Prosaccade and Antisaccade Tasks

39

Figure 4 – Experimental Setup

40

3.3.3 Saccade Coding During the laptop version, the experimenter stood behind the participant and raised a right or left index finger when the participant was suppose to gaze right or left respectively (see Figure 3), thus providing a method by which observers of the web camera videos could determine if the participant was looking in the correct direction. If the participant fixated centrally for at least 2 video frames, then made a saccade in the direction of the raised finger no sooner than 2 frames prior to the finger being raised and no later than the next trial, then the response was coded as correct. It is important to note that frame rate was variable (20 – 30 frames per second) for each video and was chosen dynamically by the web camera software; thus 2 video frames for one video would be of slightly different temporal length than two frames of another video. The fixation criteria of 2 frames is an arbitrary length of time chosen to insure participants followed instructions and returned their gaze to center after each trial. However, if they failed to fixate centrally before the next trial, their response was coded as a fixation error. Errors that were corrected before the next trial were coded as corrected errors, while those left uncorrected were coded as uncorrected. Trials in which no action was made were coded as omissions. Fixation and omission errors were excluded from the analysis of antisaccade errors and were analyzed separately. Percentage of errors was defined as: (corrected + uncorrected errors)/(# of trials) x 100. The correlation between results obtained by the main rater (LDK) and a second rater (CA), who coded videos from 20 participants, was 0.88 (p < 0.001) indicating a high reliability for coding criteria. The 20 participants rated by both coders included 10 controls and 10 patients with AD.

41

3.3.4 Statistical Analysis All statistical analysis was completed with SPSS v 16.0 (SPSS, Chicago, IL). Group comparisons were completed on each of the demographic variables and saccade variables using one-way analysis of variance (ANOVA). When the ANOVA assumption of equal variances was not met (e.g. prosaccade errors, antisaccade: error rates, corrections, fixations and omissions) the Welch’s robust tests of equality was utilized (Welch, 1947). For each participant, data from both antisaccade blocks were aggregated and analyzed as a whole. To assess antisaccade error rates in milder levels of AD, additional analysis was conducted on one subgroup with MMSE scores > 22 and another subgroup with MMSE scores > 24. The first subgroup, with MMSE scores > 22 was selected to provide a direct comparison with the study conducted by Boxer and colleagues (2006), while the second group (MMSE > 24) was selected because an MMSE score of 24 is considered a cutoff point for dementia. To assess the diagnostic capacity of antisaccade errors, uncorrected errors and fixation errors, sensitivity, specificity, positive predictive value and negative predictive value were calculated. Sensitivity and specificity calculation required binary classification of performance; therefore, antisaccade performance was categorized as impaired (two standard deviations above the NC mean) or unimpaired (under two standard deviations of the NC mean). The effect size of antisaccade error rates was calculated with Cohen’s d, which is the mean difference in antisaccade errors between the two groups, divided by the pooled standard deviation (Cohen, 1988). Values derived from the Cohen’s d test are categorized into effect sizes that are small (0.2
r < 0.5), medium (0.5
r < 0.8) and large (r  0.8) (Cohen, 1992).

42 To analyze the effect on response of same direction compared with different direction trials, each trial from the laptop version was categorized based on whether the preceding trial was in the same or in the different direction. A repeated measures ANOVA was used to analyze the differences between same compared with different trials across both groups, while a split-plot ANOVA was used to asses the differences between these two trial types within each group. This procedure was carried out for prosaccades (23 trials - first trial had no preceding trial) and for the 46 antisaccades trials (block 1 + block 2 - the first trial in each block).

3.4 Results Demographic data for both groups are summarized in Table 4, while Table 5 summarizes neuropsychology scores completed by a subset of the AD group. Performance metrics from the both the laptop-based and clinical antisaccade tasks are summarized in Table 6. Antisaccade error rates made during the clinical version were significantly correlated with performance on the laptop version of the antisaccade task (r = 0.579, p < 0.01). Patients with AD not only made significantly more errors on both the prosaccade (F(1,4.76) = 4.76, p < 0.05) and antisaccade tasks (F(1,47.6) = 24.72, p < 0.001), but they also left significantly more antisaccade errors uncorrected (F(1,29.5) = 22.3, p < 0.001) (see Figure 5). During the antisaccade task, patients made significantly more fixation errors (F(1,31.7) = 23.6, p < 0.01) and omission errors (F(1,31.4) = 8.1, p < 0.01) compared with age-match controls. Both the subgroup comprised of patients with MMSE scores > 22 (F(1,31.6) = 18.24, p < 0.001) and the group comprised of patients with MMSE scores > 24 (F(1,22.3) = 14.5, p < 0.01) made significantly more antisaccade errors than the

43 normal control group (see Table 6). In the AD group, antisaccade errors were correlated with the number of non-perseverative errors on the WCST (r = 0.562, p < 0.05) but were not correlated with any of the other neuropsychological measures including the MMSE (r = -0.078, p = 0.683) or DRS (r = -0.308, p = 0.2).

44

Table 4 - Demographics Table 4. Demographics* NC

AD

N

31

30

Sex( female)

18

11

70.5 (8.2) (50 – 86)

72.3 (9.7) (51 – 92)

Years of Education

16 (2.6) (11 – 21)

14.9 (3.3) (10 – 21)

MMSE

29 (1.1) (26 – 30)

24.5 (3.2)** (17 – 30)

Age

* Age, Years of Education and MMSE are given in mean, (SD) and (Range). ** p < 0.01

Table 5 – AD Neuropsychology Results Table 5. AD Neuropsychology Results* N

Score

Range

DRS

19

123.1 (10.3)

102 – 141

DAD (%)

20

77.3 (15.7)

43 – 100

Backwards digit span (out of 12)

14

5.9 (2.1)

2–9

Trails A (errors)

19

0.0 (0)

0–0

Trails B (errors)

18

0.9 (0.9)

0–3

Wisconsin Card Sorting Task

19

Correct Responses

41.4 (6.6)

33 – 53

Non perseverative Errors

11.1 (8.5)

0 – 27

Perseverative Errors

11.9 (6.6)

1 - 27

*Current neuropsychology data was only available for a subset of patients with AD

45

Figure 5 – Antisaccade Errors and Uncorrected Errors

Error bars represent standard error

46

Table 6 - Antisaccade Performance NC (n = 31)

AD (n = 30)*

AD > 22 (n = 22)*

AD > 24 (n = 17)*

MMSE

29 (1.1)

24.5 (3.22)

26.1 (2.0)

26.8 (1.6)

Errors

22 (18.4)

54.0 (30.3)

53.5 (30.8)

53.3 (31.1)

Uncorrected Errors

1.2 (3.4)

32.4 (36.0)

27.6 (33.4)

25.5 (31.0)

Fixation Errors

1.8 (3.8)

17.4 (17.2)

16.2 (16)

14.7 (14.5)

Omission Errors

0.4 (1.1)

3.3 (5.5)

3.2 (5.6)

4.0 (6.1)

Clinical Antisaccade Errors

28.7 (19.8)

51.5 (32.1)

49.5 (27.8)

49.4 (27.9)

*Sample size for clinical antisaccade task was less (AD: = 27, AD>22 = 21, AD> 24 = 16). Three participants were unable to participate in the clinical task because of time constraints.

47 Sensitivity and specificity, as well as the cut-off points for impairment on each of the antisaccade metrics, are outlined in Table 7. While all of the metrics provided specificities greater than 0.9, sensitivity levels were low, with the greatest level of sensitivity found in uncorrected errors (sensitivity: 0.63). Prosaccade errors were not included in sensitivity and specificity analysis as the standardized magnitude of differences between groups was medium (Cohen’s d: 0.56), indicating large amounts of overlap between the two groups. Likewise analysis of interaction between trial type (same versus different) and group (AD versus NC) was not significant suggesting both groups were equally likely to make errors on same trials relative to different trials, thus sensitivity and specificity were not calculated for same and different errors.

Table 7 - Diagnostic Capacity of Antisaccade Metrics

Cut-off

Number of Impaired

Sensitivity

Specificity

PPV

NPV

Cohen’s d

Errors

> 58.8%

15 AD, 1 NC

0.5

0.97

0.94

0.67

1.28

Uncorrected Errors

> 7.9%

19 AD, 2 NC

0.63

0.94

0.90

0.73

1.22

Fixation Errors

> 9.5%

16 AD, 2 NC

0.53

0.94

0.88

0.67

1.25

Omissions

> 2.7%

9 AD, 2 NC

0.3

0.94

0.82

0.58

0.73

Clinical Antisaccade Errors

> 68.3%

11 AD, 2 NC

0.41

0.94

0.85

0.64

0.85

Cutoff: 2 standard deviations greater than mean of the elderly control group Impaired: Those who were higher than the cut-off PPV: Positive Predictive Value NPV: Negative Predictive Value

48

Significantly more errors were made during same trials compared with different trials (groups collapsed) for the prosaccade task (F(1,59) = 68.5, p < 0.01), but not the antisaccade task (F(1,59) = 0.183, p = 0.67) (Figure 6.). An interaction between groups (AD vs. NC) and trial type (same vs. different) was not significant for either prosaccades (F(1,59) = 2.14, p = 0.149) or antisaccades (F(1,59) = 0.004, p = 0.948), indicating that the ratio of same to different errors was equal for both groups.

Figure 6 – Same Direction Errors Compared To Different Direction Errors

49

3.5 Discussion The current study examined antisaccade performance in patients with mild AD and elderly controls using both a novel laptop-based antisaccade task and a clinical antisaccade task. Error rates from the clinical version of the antisaccade task closely mirrored error rates from the laptop-based antisaccade task in both patients and elderly controls. Patients with AD made significantly more antisaccade errors than controls and their scores on the laptop-based antisaccade task were significantly correlated with their clinical antisaccade scores. Patients with mild AD either with MMSE scores greater than 22 or greater than 24, also made significantly more errors than controls. In addition to increased antisaccade error rates, patients left significantly more errors uncorrected. Furthermore, large effect sizes indicate that the mean magnitude of difference between the two groups is large and could be detected in smaller sample sizes. However, despite significant differences and large effect sizes in antisaccade performance between AD and the elderly control group, antisaccade metrics were associated with low sensitivities because almost a third of AD patients were defined as unimpaired on antisaccade metrics. In contrast, antisaccade metrics provided high specificities, as only one or two participants in the NC group that were considered impaired. While both the sensitivities and effect size of the clinical antisaccade task were less than the laptop-based version, the clinical task did yield a specificity equal to the laptop-based version. In contrast to other studies, we did not find a correlation between general measures of dementia, such as the MMSE or DRS, with antisaccade error rates in our AD sample. The only neuropsychological measure, administered to the AD group, which was

50 significantly correlated with antisaccade error rates, was non-perseverative errors on the WCST; all other measures were not correlated with antisaccade errors. When trials of the same direction relative to the previous trial (“same trials”) were compared with trials of a different direction relative the previous trial (“different trial”) we found differences in error rates between the two trial types for prosaccades but not antisaccades in both the AD and elderly control group.

3.6 Clinical Antisaccade Task Based on previous findings from Curried and colleagues (1991), we hypothesized that error rates on the clinical antisaccade task would correlate with those from the laptop-based version of the antisaccade task. We found that error rates from both versions did indeed significantly correlate with each other, suggesting the clinical version of the task may provide a simple method for tracking antisaccade performance. It is important to note that we did not find a perfect one to one match in error rates between the two tasks. One possibility is that independent of error rates some of the elderly participants were less anxiety prone to perform a task with a clinician than a computer, resulting in differences between the two tasks. Another possibility is that fewer trials during the clinical task prevented participants from regressing to their mean performance, thereby reducing potential variations with laptop-based version.

3.7 Antisaccade Errors Elevated in Mild AD We hypothesized that previously reported differences in error rates between patients with mild AD and elderly controls were mainly due to the inclusion of more severely demented patients who tend to make 100% errors on the task. To test this

51 hypothesis, we tested AD patients with MMSE scores
17 and repeated our analysis on subsets of patients with MMSE scores > 22 and greater than 24. To our knowledge, only the study conducted by Boxer and colleagues (2006) has examined antisaccade error rates in mild AD and they did not find a significant difference from elderly controls. They posited that frontal pathology is a late feature in AD and thus, patients with mild AD would not have "sufficient" pathology to be impaired on the antisaccade task (Boxer et al. 2006). Mild AD is thought to correspond with Braak and Braak’s stage 4, a stage in which neurofibrillary changes in the DLPFC are still relatively mild. During Braak and Braak stages 5 – 6, which are thought to correspond with moderate to severe AD, DLPFC pathology is more evident. It would thus be expected that persons with mild AD would not have DLPFC pathology and would not be impaired on the antisaccade task. However, using a larger sample size, we have shown that about two thirds of the patients with mild AD do in fact make significantly more errors than controls, indicating that some patients may have sufficient frontal neuropathology to show antisaccade impairments. Despite the previous finding that error rates are normal in mild AD and despite Braak and Braak’s findings suggesting that DLPFC neurofibrillary pathology is relatively mild at this stage, there is mounting evidence that 1) executive deficits occur earlier in disease onset, during a pre-AD stage called Mild Cognitive Impairment and 2) in vivo DLPFC structural changes can be detected during MCI and mild AD. For instance, using MRI and FDG-PET Mosconie and colleagues (2007) identified a significant degree of DLPFC white matter atrophy in patients with MCI that progressed into AD. Other reports suggest that AD is heterogeneous, with a subset of AD associated with pronounced

52 frontal deficits that are often confused with frontal temporal dementia (Snowden et al., 2007) but are associated with a less severe self-regulator disorder (Stopford, Showden, Thompson & Neary, 2008). Although the large sample studied by Snowden and colleagues only contained 2% of AD patients with pronounced frontal deficits, it seems likely that a continuum of DLPFC pathology may exist in AD with some patients having intermediate degrees of frontal dysfunction. It is conceivable that as sample sizes become smaller the probability of capturing the variation in frontal pathology would decrease, thus the subset of patients studied by Boxer and colleagues (2006) may have been less likely to capture this variation than a study with a larger sample, such as the current study.

3.8 Dementia Severity and Antisaccade Error Rates A significant correlation between general measures of dementia, such as the DRS or the MMSE, has been consistently reported, suggesting that error rates, and ultimately DLPFC pathology, might simply be predicted by general levels of dementia. We found that the mean antisaccade error rate of AD patients, 55%, was relatively low compared with previously reported antisaccade error rates of 50 – 80% (Table 1). Although this study was not strictly comparable to previous studies, as methodologies differed, the comparison reveals that the exclusion of more severely demented patients may have resulted in lower mean error rates relative to previous studies which did include severely demented patients.

53 It is of note that we were unable to replicate the previously reported correlations between error rates and DRS or MMSE scores within the AD group. There are several possible reasons why we failed to replicate this relationship. First, the relationship between MMSE and antisaccade error rates may have in previous studies been largely determined by including more severely demented patients who consistently perform poorly on the antisaccade task. Such patients were not included in this study. As discussed above, this suggests that antisaccade error rates, and potentially frontal neuropathology, may not reflect overall dementia severity during mild stages of AD. Furthermore, as stated above, it appears larger samples of patients are necessary to encounter frontal deficits in mild AD. Second, the heterogeneous nature of AD renders the MMSE an unreliable metric for dementia severity (Stopford et al., 2008). For instance, lower MMSE scores might reflect domain specific impairments such as language or memory, while executive functions remained preserved. Both possibilities are not mutually exclusive and could contribute to the differences between this study’s findings and previous investigations.

3.9 Antisaccade Errors On Same Relative To Different Direction Trials Common statistical analyses such as t-tests or an ANOVA make the assumption that each measurement, or trial, is independent of the others. That is, if a participant completes 40 antisaccade trials, each response is in no way associated with the previous trial or response. However, it appears that this assumption may be erroneous since a

54 given trial can evidently influence subsequent trials. To study the effect of one trial on the next, previous studies have examined the difference between the number of errors made on trials of the ‘same’ or of a ‘different’ direction than the previous trial in young healthy controls, and reported more errors on different trials in the and antisaccade tasks (Barton et al., 2006). That is, an error was more likely to be made on a trial preceded by a trial of a different direction (i.e. left then right). In this study however, both patients with AD and elderly controls made more errors on ‘same’ prosaccade trials. This discrepancy may have resulted from a difference between the probabilities of same compared with different trials between this study and previous studies. The probability that a trial would be the same was only 0.34 in this study, while in Barton's study it was roughly 0.5. It is particularly noteworthy that while there was a disparity between same compared with different trials during the prosaccade task, this was not the case during the antisaccade task. One of the main differences between antisaccades and prosaccades is the level of complexity between the two conditions. Prosaccades are reflexive and can be made with little effort, while antisaccades require much more effort and concentration. During the prosaccade task, the decreased effort and concentration required may increase a participant's ability to anticipate the direction of the next trial, while during the antisaccade task, a participant may be less able to anticipate the forthcoming event because of the increased effort and concentration required.

3.10 Prosaccade Errors The high disparity in error rates between same and different trials during the prosaccade tasks may also help to explain the finding that patients made significantly

55 more prosaccade errors than controls, a finding which contrasts with previous studies that have consistently reported otherwise. One possibility is that patients with AD have a propensity for making anticipatory saccades (Fletcher & Sharpe, 1986): thus by changing the probability of same and different trials, the chance that a patient would make an anticipatory response and ultimately an error may have also increased. Another possibility is that previous studies typically administer saccade tests in darkened rooms, whereas this study carried out testing in an illuminated room, a situation that may have offered more visual distractions that ultimately increased the error rates for patients.

3.11 Antisaccade Errors and Neuropsychology An absence of a correlation between antisaccade error rates with either digit backwards or number of Trails B errors corresponds to previously reported findings (Boxer et al., 2006). Although antisaccade error rates and WCST metrics have not been compared in patients with AD it was hypothesized that the two measures would correlate based on their association with the DLPFC. In contrast, we failed to find a correlation between perseveration errors and antisaccade errors, but a correlation between nonperseverative errors was found. Despite the association of both perseveration errors and antisaccade errors with the DLPFC their precise neural correlates within the DLPFC may not actually overlap. In addition to the DLPFC, focal lesions in the superior medial regions have also associated with increases in perseverative errors (Stuss et al., 2000). Furthermore, left DLPFC lesions seem to be associated with less perseverative errors than right (Stuss et al., 2000), while focal lesion studies have reported no such laterality differences in antisaccade error rates. Thus, there may be less overlap between the neural correlates associated with antisaccade errors and perseveration errors.

56 Despite the absence of correlation in patients with AD, studies examining antisaccade performance in Schizophrenia have reported correlations between perseverative errors and antisaccade errors (Rosse et al., 1993; Tien et al., 1996). However, unlike AD, frontal pathology is a salient and early feature of Schizophrenia and may be more likely to affect both of the neural correlates associated with antisaccade errors and perseveration errors.

57

Chapter 4 4.1 Performance Monitoring and Task Setting Deficits Antisaccade errors in AD have been attributed to inhibitory deficits; however, they may result from more general cognitive processes. Stuss and Alexander (2007) fractionated frontal deficits into three domain general component processes: energization, task monitoring and task setting. Deficits in energization manifest in a gradual worsening of performance over time and appear as an unlikely culprit for AD antisaccades error rates as patients performed consistently poorly throughout the task (Crawford et al., 2005). However, there is evidence to suggest that both impairments in task monitoring and task setting may lead to high error rates in AD. An error in task setting would manifest as an inability to form a relationship between a stimulus (peripheral target) and a response (look away). A deficit in task setting might explain why patients make many antisaccade errors: because they fail to associate a peripheral target with the correct response (look away). In contrast, a deficit in task monitoring might explain why patients with AD correct significantly less errors than controls or even patients with Frontotemporal dementia (Boxer et al., 2006; Garbutt et al., 2008). Error monitoring and task setting are associated with the right and left prefrontal cortex respectively, brain regions both critical for antisaccades and that are both associated with AD neuropathology. Thus, poor antisaccade performance may reflect an underlying deficit in several domain-general processes rather than just inhibitory control.

58

4.2 Clinical Relevance of The Antisaccade Task As the antisaccade task requires neither a verbal nor a manual response, it is ideally suited for testing elderly patients who may have motor or language impairments that could obfuscate performance on standard neuropsychological tests. The initial motivation for examining antisaccade performance in patients with AD in previous studies was to assess its ability to help diagnose AD and differentiate it from other forms of dementia. However, for several reasons, this objective was not met in those investigations. First, antisaccade performance in AD was shown to be highly variable and overlapped with the performance of normal controls, making it diagnostically insensitive and nonspecific (Shafiq-Antonacci et al., 2003). While we confirmed the antisaccade metrics provide low sensitivity in patients with mild AD, in contrast to the findings of Shafiq-Antonacci (2003) we found that they were associated with high specificity. In the present study, a high specificity indicates that if an impairment in antisaccade errors, fixation errors or uncorrected errors is detected there is an extremely high probability that the participant was diagnosed with AD. Thus if the probability is high that impairment is associated with AD, there is the potential that impairment on the antisaccade task could increase the diagnostic confidence of AD. In addition to suggesting AD pathology in the right clinical context, the task could potentially be used to differentiate patients with DLPFC pathology from those without it. For instance, patients with Semantic Dementia, a disease that is not associated with DLPFC pathology but is associated with memory deficits that overlap with AD, perform normally on the antisaccade task (Boxer et al., 2006; Garbutt et al., 2008). In addition to using antisaccade error rates to differentiate those with DLPFC functional impairment

59 those without, the number of uncorrected antisaccade errors may differentiate those with AD (60% uncorrected errors) from those with FTD (40% uncorrected errors) or Primary Nonfluent Aphasia (PNFA) (50% uncorrected errors). Patients with PNFA or FTD make significantly more errors than controls on the antisaccade task, yet unlike patients with AD, these populations correct more of their errors (Boxer et al., 2006; Garbutt et al., 2008). Thus, by interpreting antisaccade deficits as a specific type of frontal deficit the task may have more utility than as a diagnostic test for AD. A second problem is that the task traditionally has relied on cost-prohibitive eyetracking equipment that can be difficult to use and difficult to interpret. This second shortcoming, cost prohibitive eye-tracking equipment and lack of expertise in eyetracking, have been overcome in this investigation by utilizing an extremely parsimonious clinical version of the antisaccade task. Currie and colleagues (1991) reported that results obtain from the clinical version, in patients with AD, were significantly correlated with infrared oculography, a commonly used method for tracking saccades. The clinical version of the antisaccade task was incorporated into a clinical HIV dementia scale that provided greater specificity and sensitivity than the MMSE (Power et al. 1995). The HIV dementia scale was later validated for use in patients with vascular cognitive impairment further highlighting the ease of use and applicability of the clinical version of the antisaccade task (van Harten et al., 2004). In agreement with Currie and colleagues (1991), we found that results from the clinical antisaccade task significantly correlated with those found on the laptop version, further validating its use. As a compromise between traditional eye tracking techniques, such as the infrared oculography and the clinical version, we also utilized a simple web camera-based eye-

60 tracking technique. The web camera-based eye-tracking technique, employed on the laptop version of the antisaccade task, provided an additional eye-tracking method that offered greater portability, ease of use and less cost than traditional eye-tracking methods while providing greater accuracy than the clinical method. Results obtained with the web camera method, specifically the percentage of antisaccade errors, fell in the middle of previously reported results for both elderly controls and patients with AD, thus affirming its validity. Although the clinical method is the most parsimonious, the web-camera method is ideal in situations where metrics other than antisaccade error rates are beneficial for diagnosis, such as uncorrected errors with AD.

4.3 Study Limitations There were several limitations with the current study that can be distilled into to two categories: 1) eye-tracking and 2) patient characteristics. The first limitation of this study was an absence of an eye-tracking gold standard, such as infraredoculography or electrooculagraphy, to compare with results obtained from the clinical or laptop-based saccade paradigms. Despite this limitation, results obtain from both the elderly control group and the AD group correspond well with previous studies, indicating an acceptable level of validity. The second set of limitations is in regards to patient characteristics. Although all the patients in the AD group meet the criteria for possible AD, we did not have pathological data which precluded the classification of definite AD. Furthermore, the AD group is a convenience sample based on patients recruited from a tertiary care centre. It is

61 conceivable that patients referred to our clinic represent complicated AD cases and may not be an accurate representation of AD in general. Finally, this study in addition to others, have consistently found high levels of variation in antisaccade performance which further highlights the need for larger sample sizes.

4.4 Future Directions The results from this study and others offer compelling evidence for clinical adoption of the antisaccade task as an index of DLPFC function: however, there are several areas of future investigation that would strengthen the task and help to eliminate other alternative explanations for elevated antisaccade error rates. There are four areas of investigation that would help to further strengthen the link between DLPFC pathology and antisaccade errors: each will be discussed.

4.4.1 Fractionation: domain specific and domain general Focal lesion studies examining discrete cortical lesions have largely excluded cortical regions, other than the DLPFC, as being necessary for antisaccades. However, AD pathology is diffuse and may impair more than one antisaccade process, creating a situation where cerebral regions, other than the DLPFC or subprocess other than inhibitory control are causing antisaccade errors. To exclude that scenario, each antisaccade sub-process such as vector inversion and fixation, as well as domain general processes such as attention and memory, will have to be excluded as the cause of increased errors in AD. Previous studies have begun to accomplish this goal; one study found equal error rates between a no-go saccade task, in which a participant simply has to inhibit a saccade, and the antisaccade task, suggesting that inhibition and not volitional

62 eye-movements was associated with increased error rates in AD (Crawford et al., 2005). Reported visuospatial deficits in AD (Buck, Black, Behrmann, Caldwell, & Bronskill, 1997; Cummings & Cole, 2002; Tippett & Sergio, 2006) might impair patients’ ability to complete a vector inversion, the process by which a prosaccade vector is inverted to make an antisaccade, but this has not been investigated in AD. Furthermore, domain general processes such as attention and memory, which are both impaired in AD, will need to be more rigorously studied in order to be eliminated as causes for antisaccade errors. A previous study found that within a block error rates do not progressively increase suggesting that patients are not gradually forgetting instructions (Crawford et al., 2005); however, it has yet to be shown that a small number of patients are not forgetting task instructions before they have even started the task.

4.4.2 Neural Correlates: Structural and Functional Imaging in Conjunction with Pathological Data Despite extensive research into the neural correlates of the antisaccade task in healthy controls, patients with focal lesions and non-human primates, there has only been a single study to do so in patients with AD. Boxer and colleagues (2006) used voxelbased morphometry to determine if there were correlations between grey matter volume and antisaccade error rates. Their analyses revealed a significant correlation between an area ventral to the right frontal eye field: larger volume was associated with lower error rates. However, by controlling for MMSE scores, which previous studies have shown is correlated with error rates, Boxer and colleagues may have eliminated variations in error rates and ultimately in DLPFC pathology.

63 Focal lesion studies have not only implicated the DLPFC as a critical structure for antisaccades but have also implicated its connections through the anterior limb of the internal capsule down to the superior colliculi (Condy et al., 2004). This network provides a simple a prior hypothesis for future imaging studies. For instance, white matter volume or fractional anisotropy could provide data for making inferences on whether the connections between the DLPFC and the superior colliculi play a role in increased error rates in AD. In contrast, gray matter volume, blood perfusion ratios (as assessed by SPECT) and blood oxygen level dependent signal (fMRI) could be used to look at both the volume and functional indices of the DLPFC in association with antisaccade performance. However, the most direct evidence for a connection between pathology and antisaccade performance would come from pathology analysis either in vivo using the new amyloid PET ligands (Dubois et al., 2007), or using traditional ex vivo pathological techniques.

4.4.3 Large Scale Validation Of The Antisaccade Task In AD The heterogeneous nature of frontal deficits in Alzheimer’s disease makes it difficult to make brain behaviour correlations from a small sample that could be generalized to other patients with AD. Thus, fractionation and imaging techniques will have to be administered to a large sample of patients with varying levels of AD who have completed both the clinical version of the antisaccade task and an antisaccade task that utilizes more traditional eye tracking techniques. The diagnostic utility of the antisaccade task in schizophrenia has been greatly aided through a large multisite study using identical testing methods and eye-tracking techniques (Radant et al., 2007). Similar

64 scientific rigour would greatly improve the understanding of antisaccade impairments and potential DLPFC pathology in AD.

4.4.4 Same Trials Versus Different Trials The statistical assumption that a response on a trial is independent from the previous has been shown to be false in healthy controls and, as we have shown, elderly controls and patients with AD. When the numbers of trials is large and when each trial has an equal probability of being left or right and an equal probability of being same or different, the disparity between errors made on same and different trials becomes small (Barton, Goff & Manoach, 2006). However, studies examining antisaccade performance in patients with AD typically administer between 20 to 40 trials which would make it even more important to control for the probability of same and different trials as well as left and right trials. From our data we hypothesize that the difference in error rates between same and different trials is great when 1) the probability of a same trial is significantly different from the probability of a different trial, and 2) only during the prosaccade task. Altering the probability of same trials, relative to different, would provide a mechanism to test the first hypothesis. In regards to the second hypothesis, the main difference that would alter ones ability to predict the direction of the next trial between the prosaccade and antisaccade tasks would task difficulty. Thus, if the difficulty of the prosaccade task were increased and the participants’ ability to predict the direction of the next trial decreased, then the disparity between errors made on same and different trials could be expected to diminish or vanish.

65

4.5 Conclusions A progressive deficit in episodic memory is the most prominent feature of Alzheimer's disease; however, there is an increasing awareness that AD is heterogeneous and is associated with impairments in the visuospatial, executive and language domains (Stopford et al., 2008). Our findings highlight that impairments in executive functions, manifested by increased antisaccade errors, 1) occur in AD much earlier than posited by previous antisaccade studies, and 2) in mild AD antisaccade errors are not correlated with general measures of dementia such as the MMSE. The findings presented in this study provide evidence that antisaccade error rates are highly variable in AD and provide a methodological framework for future studies seeking to establish and strengthen the link between antisaccade errors and DLPFC pathology in AD.

66

References

Abel, L. A., & Douglas, J. (2007). Effects of age on latency and error generation in internally mediated saccades. Neurobiol Aging, 28(4), 627-637. Abel, L. A., Unverzagt, F., & Yee, R. D. (2002). Effects of stimulus predictability and interstimulus gap on saccades in Alzheimer's disease. Dement Geriatr Cogn Disord, 13(4), 235-243. Babins, L., Slater, M. E., Whitehead, V., & Chertkow, H. (2007). Can an 18-point clockdrawing scoring system predict dementia in elderly individuals with mild cognitive impairment? Journal of Clinical and Experimental Neuropsychology, 30(2), 1-14. Barton, J.J., Goff, D.C. & Manoach, D.S. (2006). The inter-trial effects of stimulus and saccadic direction on prosaccades and antisaccades, in controls and schizophrenia patients. Exp Brain Res, 174, 487-1498. Barnea-Goraly, N., Menon, V., Eckert, M., Tamm, L., Bammer, R., Karchemskiy, A., et al. (2005). White matter development during childhood and adolescence: a crosssectional diffusion tensor imaging study. Cereb Cortex, 15(12), 1848-1854. Behl, P., Stefurak, T. L., & Black, S. E. (2005). Progress in clinical neurosciences: cognitive markers of progression in Alzheimer's disease. The Canadian journal of neurological sciences, 32 (2), 140-151

67 Boxer, A. L., Garbutt, S., Rankin, K. P., Hellmuth, J., Neuhaus, J., Miller, B. L., et al. (2006). Medial versus lateral frontal lobe contributions to voluntary saccade control as revealed by the study of patients with frontal lobe degeneration. J Neurosci, 26(23), 6354-6363. Braak, H., & Braak, E. (1991). Neuropathological stageing of Alzheimer-related changes. Acta Neuropathol (Berl), 82(4), 239-259. Buck, B. H., Black S. E., Behrmann, M, Caldwell, C., & Bronskill, M. J. Spatial- and object-based attentional deficits in Alzheimer's disease. Relationship to HMPAOSPECT measures of parietal perfusion. Brain (1997) vol. 120 ( Pt 7) pp. 1229-44 Cohen, J. (1998). Statistical Power Analysis for the Behavioral Sciences, ed 2. Hillsdale, Erlbaum. Cohen, J. (1992). A Power Primer. Psychological Bulletin, 112, 155-159. Condy, C., Rivaud-Pechoux, S., Ostendorf, F., Ploner, C. J., & Gaymard, B. (2004). Neural substrate of antisaccades: role of subcortical structures. Neurology, 63(9), 1571-1578. Connolly, J. D., Goodale, M. A., Desouza, J. F., Menon, R. S., & Vilis, T. (2000). A comparison of frontoparietal fMRI activation during anti-saccades and antipointing. J Neurophysiol, 84(3), 1645-1655.

68 Crawford, T. J., Higham, S., Renvoize, T., Patel, J., Dale, M., Suriya, A., et al. (2005). Inhibitory control of saccadic eye movements and cognitive impairment in Alzheimer's disease. Biol Psychiatry, 57(9), 1052-1060. Cummings, J. L., & Cole, G. (2002). Alzheimer disease. Jama, 287(18), 2335-2338. Currie, J., Ramsden, B., McArthur, C., & Maruff, P. (1991). Validation of a clinical antisaccadic eye movement test in the assessment of dementia. Arch Neurol, 48(6), 644-648. DeSouza, J. F., Menon, R. S., & Everling, S. (2003). Preparatory set associated with prosaccades and anti-saccades in humans investigated with event-related FMRI. J Neurophysiol, 89(2), 1016-1023. Dubois, B., Feldman, H. H., Jacova, C., Dekosky, S. T., Barberger-Gateau, P., Cummings, J., et al. (2007). Research criteria for the diagnosis of Alzheimer's disease: revising the NINCDS-ADRDA criteria. Lancet Neurol, 6(8), 734-746. Ettinger, U., Antonova, E., Crawford, T.J., Mitterschiffthaler, M.T. et al. (2005). Structural neural correlates of prosaccade and antisaccade eye movements in healthy humans. Neuroimage, 24(2), 487-94 Everling, S., Spantekow, A., Krappmann, P., & Flohr, H. (1998). Event-related potentials associated with correct and incorrect responses in a cued antisaccade task. Exp Brain Res, 118(1), 27-34. Feldman, H., Sauter, A., Donald, A., Gelinas, I., Gauthier, S. et al. (2001). The disability assessment for dementia scale: a 12-month study of functional ability in mild to

69 moderate severity Alzheimer disease. Alzheimer disease and associated disorders, 15, 89-95. Ferri, C. P., Prince, M., Brayne, C., Brodaty, H., Fratiglioni, L., Ganguli, M., et al. (2005). Global prevalence of dementia: a Delphi consensus study. Lancet, 366(9503), 2112-2117. Fletcher, W. A., & Sharpe, J. A. (1986). Saccadic eye movement dysfunction in Alzheimer's disease. Ann Neurol, 20(4), 464-471. Ford, K. A., Goltz, H. C., Brown, M. R., & Everling, S. (2005). Neural processes associated with antisaccade task performance investigated with event-related FMRI. J Neurophysiol, 94(1), 429-440. Garbutt, S., Matlin, A., Hellmuth, J., et al. (2008). Oculomotor function in frontotemporal lobar degeneration, related disorders and Alzheimer's disease. Brain, 131, 12681281. Garrido, G. E. J., Furuie, S. S., Buchpiguel, C. A., Bottino, C. M. C., Almeida, P., Cid, C. G., et al. (2002). Relation between medial temporal atrophy and functional brain activity during memory processing in Alzheimer's disease: a combined MRI and SPECT study. Journal of Neurology Neurosurgery and Psychiatry, 73(5), 508516. Gauthier, S., Gélinas, I., Gauthier, L. (1997). Functional disability in Alzheimer's disease. International psychogeriatrics, 9(Suppl 1), 163-165.

70 Gaymard, B., Francois, C., Ploner, C. J., Condy, C., & Rivaud-Pechoux, S. (2003). A direct prefrontotectal tract against distractibility in the human brain. Ann Neurol, 53(4), 542-545. Gaymard, B., Ploner, C. J., Rivaud-Pechoux, S., & Pierrot-Deseilligny, C. (1999). The frontal eye field is involved in spatial short-term memory but not in reflexive saccade inhibition. Experimental Brain Research, 129(2), 288-301. Giedd, J. N., Blumenthal, J., Jeffries, N. O., Castellanos, F. X., Liu, H., Zijdenbos, A., et al. (1999). Brain development during childhood and adolescence: a longitudinal MRI study. Nat Neurosci, 2(10), 861-863. Good, C. D., Johnsrude, I. S., Ashburner, J., Henson, R. N., Friston, K. J., & Frackowiak, R. S. (2001). A voxel-based morphometric study of ageing in 465 normal adult human brains. Neuroimage, 14(1 Pt 1), 21-36. Grady, C. L., McIntosh, A. R., Beig, S., Keightley, M. L., Burian, H., & Black, S. E. (2003). Evidence from functional neuroimaging of a compensatory prefrontal network in Alzheimer's disease. Journal of Neuroscience, 23(3), 986-993. Guitton, D., Buchtel, H. A. & Douglas, R. M. (1985). Frontal lobe lesions in man cause difficulties in suppressing reflexive glances and in generating goal-directed saccades. Experimental brain research, 58(3), 455-72. Hallett, P. E. (1978). Primary and secondary saccades to goals defined by instructions. Vision Res, 18(10), 1279-1296.

71 Hutton, S. B., & Ettinger, U. (2006). The antisaccade task as a research tool in psychopathology: a critical review. Psychophysiology, 43(3), 302-313. Johnson JK, Head E, Kim R, Starr A & Cotman CW. Clinical and pathological evidence for a frontal variant of Alzheimer disease. Arch Neurol. 1999;56:1233-1239. Josephs, K. A., Petersen, R. C., Knopman, D. S., Boeve, B. F., Whitwell, J. L., Duffy, J. R., et al. (2006). Clinicopathologic analysis of frontotemporal and corticobasal degenerations and PSP. Neurology, 66(1), 41-48. Kertesz, A., Martinez-Lage, P., Davidson, W., & Munoz, D. G. (2000). The corticobasal degeneration syndrome overlaps progressive aphasia and frontotemporal dementia. Neurology, 55(9), 1368-1375. Kimmig, H., Greenlee, M. W., Gondan, M., Schira, M., Kassubek, J., & Mergner, T. (2001). Relationship between saccadic eye movements and cortical activity as measured by fMRI: quantitative and qualitative aspects. Exp Brain Res, 141(2), 184-194. Klein, C., Foerster, F., Hartnegg, K., & Fischer, B. (2005). Lifespan development of proand anti-saccades: multiple regression models for point estimates. Brain Res Dev Brain Res, 160(2), 113-123. Masellis, M., Momeni, P., Meschino, W., Heffner, R., Elder, J., et al. (2006) Novel splicing mutation in the progranulin gene causing familial corticobasal syndrome. Brain, 129(Pt 11), 3115-3123

72 McDowell, J. E., & Clementz, B. A. (2001). Behavioral and brain imaging studies of saccadic performance in schizophrenia. Biol Psychol, 57(1-3), 5-22. McDowell, J. E., Kissler, J. M., Berg, P., Dyckman, K. A., Gao, Y., Rockstroh, B., et al. (2005). Electroencephalography/magnetoencephalography study of cortical activities preceding prosaccades and antisaccades. Neuroreport, 16(7), 663-668. Medendorp, W. P., Goltz, H. C., & Vilis, T. (2005). Remapping the remembered target location for anti-saccades in human posterior parietal cortex. J Neurophysiol, 94(1), 734-740. Medina, D., DeToledo-Morrell, L., Urresta, F., Gabrieli, J. D., Moseley, M., Fleischman, D., et al. (2006). White matter changes in mild cognitive impairment and AD: A diffusion tensor imaging study. Neurobiol Aging, 27(5), 663-672. Meyniel, C., Rivaud-Pechoux, S., Damier, P., & Gaymard, B. (2005). Saccade impairments in patients with fronto-temporal dementia. J Neurol Neurosurg Psychiatry, 76(11), 1581-1584. Milner, B. (1967). Effects of Different Brain Lesions on Card Sorting. Archives of Neurology, 9, 90-100. Mosconi, L., Brys, M., Glodzik-Sobanska, L., De Santi, S., Rusinek, Henry, de Leon, M. J. (2007). Early detection of Alzheimer's disease using neuroimaging. Experimental Gerontology, 42(1-2), 129-38.

73 Mosimann, U. P., Muri, R. M., Burn, D. J., Felblinger, J., O'Brien, J. T., & McKeith, I. G. (2005). Saccadic eye movement changes in Parkinson's disease dementia and dementia with Lewy bodies. Brain, 128(Pt 6), 1267-1276. Munoz, D. P., Broughton, J. R., Goldring, J. E., & Armstrong, I. T. (1998). Age-related performance of human subjects on saccadic eye movement tasks. Exp Brain Res, 121(4), 391-400. Munoz, D. P., & Everling, S. (2004). Look away: the anti-saccade task and the voluntary control of eye movement. Nat Rev Neurosci, 5(3), 218-228. Neary, D., Snowden, J. S., Gustafson, L., Passant, U., Stuss, D., Black, S., et al. (1998). Frontotemporal lobar degeneration: a consensus on clinical diagnostic criteria. Neurology, 51(6), 1546-1554. Nyffeler, T., Rivaud-Pechoux, S., Pierrot-Deseilligny, C., Diallo, R., & Gaymard, B. (2007). Visual vector inversion in the posterior parietal cortex. Neuroreport, 18(9), 917-920. O'Driscoll, G. A., Alpert, N. M., Matthysse, S. W., Levy, D. L., Rauch, S. L., & Holzman, P. S. (1995). Functional neuroanatomy of antisaccade eye movements investigated with positron emission tomography. Proc Natl Acad Sci U S A, 92(3), 925-929. O'Keeffe, F. M., Murray, B., Coen, R. F., Dockree, P. M., Bellgrove, M. A., Garavan, H., et al. (2007). Loss of insight in frontotemporal dementia, corticobasal degeneration and progressive supranuclear palsy. Brain, 130(Pt 3), 753-764.

74 Olincy, A., Ross, R. G., Youngd, D. A., & Freedman, R. (1997). Age diminishes performance on an antisaccade eye movement task. Neurobiol Aging, 18(5), 483489. Perry, R. J., & Hodges, J. R. (1999). Attention and executive deficits in Alzheimer's disease. A critical review. Brain, 122 ( Pt 3), 383-404. Pfefferbaum, A., & Sullivan, E. V. (2003). Increased brain white matter diffusivity in normal adult aging: relationship to anisotropy and partial voluming. Magn Reson Med, 49(5), 953-961. Pierrot-Deseilligny, C., Rivaud, S., Gaymard, B., & Agid, Y. (1991). Cortical control of reflexive visually-guided saccades. Brain, 114 ( Pt 3), 1473-1485. Pike, K. E., Savage, G., Villemagne, V. L., Ng, S., Moss, S. A. et al. (2007). Betaamyloid imaging and memory in non-demented individuals: evidence for preclinical Alzheimer's disease. Brain, 130(11), 2837-2844. Ploner, C. J., Gaymard, B. M., Rivaud-Pechoux, S., & Pierrot-Deseilligny, C. (2005). The prefrontal substrate of reflexive saccade inhibition in humans. Biol Psychiatry, 57(10), 1159-1165. Power, C., Selnes, O. A., Grim, J. A., & McArthur, J. C. (1995). HIV Dementia Scale: a rapid screening test. J Acquir Immune Defic Syndr Hum Retrovirol, 8(3), 273-278. Radant, A. D., Dobie, D. J., Calkins, M. E., Olincy, A., Braff, D. L., Cadenhead, K. S., et al. (2007). Successful multi-site measurement of antisaccade performance deficits in schizophrenia. Schizophr Res, 89(1-3), 320-329.

75 Raemaekers, M., Vink, M., van den Heuvel, M. P., Kahn, R. S., & Ramsey, N. F. (2005). Brain activation related to retrosaccades in saccade experiments. Neuroreport, 16(10), 1043-1047. Raemaekers, M., Vink, M., van den Heuvel, M. P., Kahn, R. S., & Ramsey, N. F. (2006). Effects of aging on BOLD fMRI during prosaccades and antisaccades. J Cogn Neurosci, 18(4), 594-603. Reuter, B, Philipp, A, M, & Kathmann N. Effects of switching between leftward and rightward pro- and antisaccades. Biological psychology (2006) vol. 72 (1) pp. 8895 Rivaud-Pechoux, S., Vidailhet, M., Brandel, J. P., & Gaymard, B. (2007). Mixing proand antisaccades in patients with parkinsonian syndromes. Brain, 130, 256-264. Shafiq-Antonacci, R., Maruff, P., Masters, C., & Currie, J. (2003). Spectrum of saccade system function in Alzheimer disease. Arch Neurol, 60(9), 1272-1278. Snowden, J., Neary, D., & Mann, D. (2007). Frontotemporal lobar degeneration: clinical and pathological relationships. Acta Neuropathol (Berl), 114(1), 31-38. Snowden, J. S., Stopford, C. L., Camille, L., Thompson, J. C., Davidson, J. C., et al. (2007). Cognitive phenotypes in Alzheimer's disease and genetic risk. Cortex; a journal devoted to the study of the nervous system and behavior, 43(7) 835-45. Stopford C. L., Snowden, J. S., Thompson, J. C. & Neary, D. (2008). Variability in cognitive presentation of Alzheimer's disease. Cortex; a journal devoted to the study of the nervous system and behavior, vol. 44 (2), 185-95.

76 Stuss, D. T., & Alexander, M. P. (2007). Is there a dysexecutive syndrome? Philos Trans R Soc Lond B Biol Sci, 362(1481), 901-915. Tatler, B. W., & Hutton, S. B. Trial by trial effects in the antisaccade task. Experimental brain research (2007) vol. 179 (3) pp. 387-96 Tippett, W. J. & Sergio, L. E. Visuomotor integration is impaired in early stage Alzheimer's disease. Brain Res (2006) vol. 1102 (1) pp. 92-102 van Harten, B., Courant, M. N., Scheltens, P., & Weinstein, H. C. (2004). Validation of the HIV Dementia Scale in an elderly cohort of patients with subcortical cognitive impairment caused by subcortical ischaemic vascular disease or a normal pressure hydrocephalus. Dement Geriatr Cogn Disord, 18(1), 109-114. Vidailhet, M., Rivaud, S., Gouider-Khouja, N., Pillon, B., Bonnet, A. M., Gaymard, B., et al. (1994). Eye movements in parkinsonian syndromes. Ann Neurol, 35(4), 420426. Welch, B. L. (1947). The Generalization of “student’s” problem when several different population variances are involved. Biometrika, 34, 28-35.