Impact of Antipsychotic Treatment on Attention and Motor Learning Systems in First-Episode Schizophrenia

Abstract

Background: Antipsychotic medications have established clinical benefit, but there are few neuroimaging studies before and after initiating antipsychotic medication to assess drug influence on brain circuitry. Attention and motor learning tasks are promising approaches for examining treatment-related changes in frontostriatal systems. Methods: Twenty-one unmedicated first-episode schizophrenia patients (14 antipsychotic-naïve) participated in functional imaging studies while performing visual attention (prosaccades) and motor learning tasks (predictive saccades). Posttreatment testing was completed in 14 patients after 4–6 weeks of antipsychotic treatment. Matched healthy controls were studied in parallel. Results: Pretreatment, patients had reduced activation in the dorsal neocortical visual attention network. Activation deficits were significantly reduced posttreatment. Higher medication dose was associated with greater caudate activation at follow-up. For the motor learning task, patients’ dorsolateral prefrontal cortex (DLPFC) was unimpaired prior to treatment but showed significantly reduced activation after treatment. Conclusion: Impairments in dorsal cortical attention networks are present in untreated first-episode schizophrenia patients. These impairments are reduced after antipsychotic treatment, suggesting a beneficial effect on neural systems for attention. Treatment-emergent decreases in DLPFC activation observed for the motor learning task are consistent with other clinical and preclinical evidence suggesting that antipsychotics can have adverse effects on prefrontal function.

Introduction

Neurophysiological abnormalities associated with acute psychosis in schizophrenia remain poorly understood. Studies of functional brain systems in acutely ill first-episode patients before and after treatment are useful for elucidating neural system substrates of the disorder, as well as mechanisms of therapeutic benefits of antipsychotic drugs. Changes in frontostriatal circuitry are believed to be important systems-level mechanisms of both illness^1-4 and the efficacy of antipsychotic medication,^5-9 but relatively little empirical clinical evidence exists to validate this model. In addition, available clinical data suggest that both beneficial and adverse effects on brain and behavioral systems occur with treatment.^10-14

Functional imaging studies of frontostriatal circuits using task-based approaches are a promising strategy for assessing their integrity across states of illness. Studies of attention and motor learning systems using eye movement tasks are particularly useful in this regard, because the neural circuitry supporting behavioral performance is well-delineated from prior work in humans and nonhuman primates,^15-18 and behavioral deficits in schizophrenia are well established in laboratory studies.^12,19,20

Automatic visual attentional processes have been examined with prosaccade tasks, in which a visual target unexpectedly relocates, eliciting a shift in attention and gaze. Untreated first-episode schizophrenia patients show abnormally rapid shifts of gaze^19,20 and this abnormality is reduced after antipsychotic treatment.¹⁹ These exogenously elicited shifts of attention are supported by the frontal and supplementary eye fields (FEF, SEF) and the intraparietal sulcus (IPS).²¹ These dorsal cortical areas influence saccades via direct excitatory projections to the superior colliculi and indirectly via the frontal-caudate-nigra-collicular pathway. First-episode schizophrenia patients performing prosaccades have shown pretreatment alterations in this dorsal cortical attention system that improved after antipsychotic treatment,¹⁰ consistent with findings from behavioral studies.¹⁹

Frontostriatal systems for motor learning have been investigated in schizophrenia with the predictive saccade task, in which a simple target relocation sequence (alternating between 2 locations at a fixed time interval) facilitates rapid learning and elicits saccades with reduced reaction times. Such predictive saccades are generated on the basis of endogenous representations of the target’s timing and location. Relative to prosaccades, predictive saccades are more dependent upon striatal motor learning and timing systems, presupplementary eye fields (preSEF) supporting increased response planning demands, and dorsolateral prefrontal and hippocampal cortex, supporting spatial learning and memory aspects of task performance.¹⁷ After 4–6 weeks of antipsychotic treatment, first-episode patients make inaccurate/hypometric predictive saccades.¹² Functional imaging studies have not been conducted to identify the neural substrate of this treatment-emergent abnormality. Prior task-based imaging studies in first-episode patients have found treatment-emergent reductions in DLFPC and striatum activation,^10,22,23 so alterations to striatal and frontal systems may underlie the behavioral changes observed for predictive saccades after antipsychotic treatment.

The aim of this study was to examine neural activation before and after antipsychotic treatment in first-episode schizophrenia patients performing pro- and predictive saccades. We hypothesized that pretreatment, patients would show reduced activation in dorsal cortical regions associated with visual attention processes, and show improvement in these deficits after treatment. We further hypothesized that for the predictive saccade task, patients would show reduced activation in dorsal prefrontal cortex and striatum posttreatment.

Methods

Participants

First-episode patients with schizophrenia spectrum disorders (schizophrenia n = 17, schizoaffective disorder depressed n = 3, schizophreniform disorder n = 1) were recruited in- or outpatient at the University of Illinois Medical Center. Consensus diagnoses including review of Structured Clinical Interview for DSM-IV²⁴ results and all available clinical information. A total of 14 patients were antipsychotic naïve. Median lifetime antipsychotic exposure for the remaining 7 patients was 2 weeks; range was from 1 day to 4 weeks. The antipsychotic exposure was >6 weeks prior to pretreatment scanning except for 1, who had poor compliance with a brief trial of antipsychotic treatment and was an inpatient scanned after being medication-free for 5 days.

A comparison group of 22 demographically matched healthy participants (table 1) without known history of Axis I disorders or history of psychotic or mood disorder in first-degree relatives was recruited via community advertisement. All subjects had no known systematic or neurological disease, no head trauma with loss of consciousness > 10 minutes, no coffee, tea or cigarettes 2 hours before testing, and no history of alcohol or drug dependence. Two patients reported past cannabis abuse more than 3 months before testing. The study was approved by the university Institutional Review Board, and all participants provided written informed consent.

Table 1.

Mean (SD) Characteristics of Participants, Assessed With t tests (Sex Evaluated With Chi Square), All Nonsignificant

	Schizophrenia (n = 21)	Healthy (n = 21)
	(Posttreatment, n = 14)	(Re-scanned, n = 12)
Age	23.9 (7.9)	24.7 (4.6)
Sex (male/female)	16/5	10/12
Premorbid IQ 25	97.4 (11.7)	99.6 (8.7)
Parental SES 26	35.8 (14.2)	41.2 (11.2)
Scorable trials	188 (85)	198 (109)
PANSS 27 positive	Pretreatment: 24.0 (4.4)
	Posttreatment: 15.6 (3.8)*
PANSS negative	Pretreatment: 18.9 (5.1)
	Posttreatment: 17.6 (5.8)
PANSS total	Pretreatment: 85.0 (14.6)
	Posttreatment: 68.1 (15.0)*
HAM-D 28	Pretreatment: 29.0 (11.6)
	Posttreatment: 17.3 (8.3)*

Notes: SES, socioeconomic status; PANSS, Positive and Negative Syndrome Scale; HAM-D, Hamilton Depression Rating Scale.

*Significantly reduced scores after treatment (P < .05).

Fourteen patients were re-scanned after a mean of 34.9 (SD = 8.1) days of antipsychotic treatment. The medication of choice was risperidone, dosed per clinician judgment; alternatives were available per clinician judgment. Twelve patients were treated with risperidone (mean daily dose at follow-up = 2.5 mg [SD = 1.6]), and 2 with 15 mg aripiprazole, leading to significant clinical improvement (table 1). Correlations with drug dose were only examined in the risperidone-treated group. The 7 patients who completed only baseline studies either refused re-testing or discontinued treatment prior to follow-up. There were no pretreatment performance differences on the tasks when comparing patients with posttreatment data vs those without. However, relative to patients not re-scanned, patients scanned posttreatment had less severe pretreatment positive symptoms (PANSS²⁷ positive scale scores were 22.1 vs 27.7; P < .05) and were younger (21.4 vs 28.8 mean years of age; P < .05). Twelve healthy participants were re-scanned at an interval matching the patients.

Functional Magnetic Resonance Imaging Task

The task followed that of Simo.¹⁷ Participants tracked a target, a 0.5° diameter white dot, appearing against a dark grey background as it changed location along the horizontal meridian. Blocks consisted of 20 target relocations. In 8 prosaccade blocks, the target shifted unpredictably 3° to the left or right of its location. In 8 predictive saccade blocks, the target moved back and forth between the same 2 locations separated by 3° for all 20 trials. Each predictive saccade block began with the final target location of the preceding prosaccade trial. The interstimulus interval for trials on both tasks was 1 second. Leftward and rightward saccades were balanced for both tasks. Saccade blocks alternated in a fixed order (prosaccade, predictive, prosaccade, and predictive) without explicit cueing for the start of each block. A fixation condition consisting of a white central crosshair was presented for 20 seconds at the beginning and end of the task, and for 30 seconds after every 4 saccade task blocks.

Eye movements were recorded by a limbus tracker (Cambridge Research Systems, UK) mounted on the headcoil and digitized at 500 Hz. Saccade latencies were measured in relation to trial onset. Average prosaccade latency was the performance variable of interest for the prosaccade task. Motor learning performance was measured by calculating proportion of anticipatory saccades, defined as saccades with latencies shorter than 110 ms, a cutoff in range of prior work.^12,29

Image Acquisition and Analysis

Functional magnetic resonance imaging (fMRI) studies were conducted using a 3.0 Tesla scanner (Signa VHi, General Electric Medical Systems, Waukesha, WI) with a gradient echo, echo-planar sequence (epiRT, 25 5-mm axial slices, skip 1 mm; TR = 2000 ms, TE = 25 ms, flip angle = 90, matrix = 64 × 64, FOV = 20 cm²). Anatomic images were acquired (3D spoiled gradient recalled, 1.5-mm thick axial slices) for co-registration. Time series data were normalized, motion-corrected, and time shifted.

Individual activation maps were constructed by conducting voxelwise t tests of the differences between the BOLD signal for each saccade task vs fixation. This yielded 1 prosaccade and 1 predictive saccade statistical activation map for each subject. These were transformed to Fisher z′ statistics, warped into Talairach space, resampled (3 × 3 × 3 mm), and spatially filtered (5 mm FWHM). Spatial standardization and all subsequent image analyses were conducted with Analysis of Functional NeuroImages.³⁰ To identify alterations pretreatment, between-group t tests for each saccade task were conducted with all baseline scans. To provide the most robust, direct test of change between baseline and follow-up scans, pairwise t tests were conducted for each group. Secondarily, t tests were conducted comparing patients’ and controls’ follow-up scans. For all comparisons, high-probability white matter voxels were excluded, and a contiguity threshold was applied, requiring 40 contiguous voxels at an individual voxel threshold of P = .025, providing a family-wise significance level of P < .05 (per AFNI’s AlphaSim). Given expected effects of antipsychotic treatment in the caudate nucleus, activation in voxels of left and right caudate was extracted separately for each patient at each scan and task (supplementary figure 1) for correlation with risperidone dose, task performance, and clinical ratings.

Results

Prosaccade Task

Task Performance.

There were no significant prosaccade latency differences between groups at baseline, nor over time (supplementary figure 2).

Brain Activation.

Prior to treatment, schizophrenia patients displayed reduced activation relative to healthy participants throughout the neocortical network associated with attentional control of saccades, including frontal cortex (FEF, SEF, preSEF, superior frontal gyrus), left IPS, and regions known to be anatomically and functionally connected to this neocortical attention network, including left lentiform nucleus, posterior cingulate, and bilateral insula. Schizophrenia patients also had reduced activation relative to controls in higher-order visual cortex, including posterior superior temporal sulci, right middle temporal area (V5), right supramarginal gyrus, and bilateral, largely inferior (lingual) occipital cortex (table 2 and figure 1A).

Table 2.

Reduced Activation in Patients Pretreatment Relative to Controls During Prosaccades (No Increases Found)

	Cluster Size (# Voxels)	Peak t	Center of Mass
			X	Y	Z
R frontal eye fielda	335b	3.84	48	−17	31
L frontal eye fielda	279c	4.16	−48	−13	41
L supplementary eye fielda	47	3.09	−7	−10	62
Bilateral presupplementary eye fields	113d	3.88	−8	22	55
Bilateral superior frontal gyrus (BA 9 and 10)a	555d	3.91	5	42	35
L intraparietal sulcusa	90e	3.84	−16	−70	44
L Posterior cingulate (18)	65e	3.80	−11	−54	17
L lentiform nucleusa	85	4.13	−26	−7	10
R insula	145b	4.13	31	−1	−10
L insula	196c	3.98	−45	9	−7
L posterior superior temporal sulcus	133	3.46	−49	−57	9
R posterior superior temporal sulcusa	142b	3.62	42	−39	−1
R anterior middle temporal cortexa	75b	3.91	39	2	−34
R supramarginal gyrus	45	4.14	41	−60	23
Bilateral inferior occipital cortex (BA 17, 18, 19)a	239	4.05	13	−73	−16
	72	3.89	−25	−64	−17

Notes: Four of the significant clusters from the analysis extended over multiple brain structures. These larger clusters were parsed into different brain structures in accordance with the oculomotor regions of interest map used in Keedy et al. ³¹ Footnotes indicate which of these 4 large clusters a reported region was part of, if any. Negative values are left, posterior, and inferior to anterior commissure for X, Y, and Z axes, respectively.

^aRegion where reduced activation was found for patients’ baseline activation when restricted to only those with follow-up scans (see supplementary table 3).

^bPortion of a large cluster that totaled 697 voxels.

^cPortion of a large cluster that totaled 475 voxels.

^dPortion of a large cluster that totaled 668 voxels.

^ePortion of a large cluster that totaled 155 voxels.

Fig. 1.

Right anatomical side shown on left. Pretreatment, patients had only reduced activation relative to controls during prosaccades (row A) and predictive saccades (row C). After treatment, patients had increased activation relative to pretreatment for both tasks (row B prosaccade; row D predictive saccade). For the predictive saccade task, patients also had reduced activation (green voxels) in right dorsolateral prefrontal cortex after treatment. Slices above (positive) or below (negative) anterior commissure (mm): (A) 56, 25, 10, −7; (B) 59, 44, 18, 0; (C) 56, 41, 8, 0; (D) 33, 11, −15.

Pretreatment to posttreatment, schizophrenia patients had significantly increased activation in the SEF, IPS, posterior superior temporal cortex, and occipital cortex (table 3 and figure 1B). No significant increases were seen in any region in healthy participants (supplementary table 1 and supplementary figure 3), and no activation deficits were found in patients’ posttreatment scans relative to the second scan of healthy participants.

Table 3.

Significantly Changed Activation Within the Schizophrenia Group After Treatment Compared With Pretreatment for Each Task

	Cluster Size (# Voxels)	Peak t	Center of Mass
			X	Y	Z
Prosaccade task (posttreatment > pretreatment)
R intraparietal sulcus	63	4.06	16	−64	43
R supplementary eye field	47	4.64	13	−4	58
L occipital (BA 18, 19)	167	4.60	−31	−81	6
R posterior superior temporal sulcus	42	4.81	40	−62	18
Predictive saccade task (posttreatment > pretreatment)
Bilateral inferior occipital (BA 17, 18)	62	4.12	4	−72	−17
L Intraparietal sulcus	45	4.43	−35	−52	32
Predictive saccade task (posttreatment < pretreatment)
R dorsolateral prefrontal (BA 10, 46)	41	3.50	39	49	9

Activation Correlations.

In risperidone-treated patients, higher posttreatment bilateral caudate activation was associated with higher risperidone dose (r = .60, P < .05, figure 2A). No other associations among performance, dose, or symptom ratings were significant.

Fig. 2.

(A) Higher risperidone dose was associated with higher activation in bilateral caudate posttreatment for prosaccades (r = .60, P < .05; red squares) and at a trend level for predictive saccades (r = .51, P = .11, green circles). (B) Mean (1 SD) proportion of anticipatory saccades during the predictive saccade task was reduced in patients prior to treatment, but improved to normal after treatment (*P < .05). Baseline differences were found for whole sample, but graph depicts data only from participants re-scanned (14 patients, 12 controls). (C and D) Activation changes (posttreatment mean activation minus pretreatment mean activation) correlated at trend levels of significance with changes in predictive saccade task performance. Fewer anticipatory saccades after treatment associated with greater reduction in right dorsolateral prefrontal cortex activation (C; r = .51, P = .07), and with greater increases in right caudate (D; r = .50, P = .08). These associations did not change when controlling for antipsychotic dose, shown for each case as chlorpromazine equivalents³².

Predictive Saccade Task

Task Performance.

Before treatment, patients made a lower proportion of anticipatory saccades during the predictive saccade task relative to healthy participants (t = 2.3, P < .05; effect was similar if restricted to only patients with follow-up data: t = 2.1, P < .05). After treatment, patients had a significant increase in proportion of anticipatory saccades (t = 2.4, P < .05) and no longer differed from healthy controls, who did not change over time (figure 2B).

Brain Activation.

Before treatment, schizophrenia patients displayed reduced activation relative to healthy participants during the predictive saccade task in frontoparietal attention systems (FEF, preSEF, IPS; table 4 and figure 1C). There were also reductions in medial-anterior prefrontal cortex and in extrastriate visual areas, parahippocampus, and anterior temporal cortex.

Pretreatement to posttreatment, patients had significantly increased activation in left IPS and occipital cortex, and reduced activation in right dorsolateral prefrontal cortex (rDLPFC; figure 1D and table 3). Secondary analyses comparing patients’ posttreatment scans to healthy participants’ second scan indicated reduced activation in patients in 2 regions: posterior superior temporal sulcus and anterior temporal cortex.

Activation Correlations.

A trend was found for an association between higher right caudate activation posttreatment and higher risperidone dose (r = .51, P = .11), paralleling effects in the prosaccade task (figure 2A). Increased right caudate activation relative to pretreatment had a trend relationship to increased rate of predictive saccades (r = .50, P = .08; figure 2D). Dose was not related to performance.

To assess whether activation changes pretreatment to posttreatment were related to performance changes, activation change scores were calculated for each patient in each of the 3 brain regions showing significant pretreatment to posttreatment alteration: left IPS, occipital cortex, and rDLPFC (see supplementary methods). All but 1 patient displayed less activation after treatment relative to pretreatment in rDLPFC, and those showing the greatest decline tended to show the least performance improvement (r = .51, P = .07, figure 2C). No other significant correlations were found among performance, dose, or symptom ratings.

Selection Bias Checks.

Baseline brain activation was compared between only those patients and controls who were scanned at follow-up. Results were consistent with that of the whole baseline sample (asterisked regions in tables 2 and 4; supplementary tables 3 and 4). Next, t tests comparing baseline brain activation in the 14 patients followed up with the 7 patients not followed up yielded no significant differences for the prosaccade task, nor in any brain region reported as different from healthy controls for the predictive saccade task (supplementary figure 5).

Table 4.

Reduced Activation in Patients Pretreatment Relative to Controls During the Predictive Saccade Task (No Increases Found)

	Cluster Size (# Voxels)	Peak t	Center of Mass
			X	Y	Z
L frontal eye fielda	204	4.22	−43	−15	35
	196	4.19	−37	11	31
	45	4.15	−23	−11	47
R frontal eye fielda	122	4.01	44	−20	35
Bilateral presupplementary eye field	112b	4.30	−9	15	52
L intraparietal sulcus	79	4.16	−25	−64	47
	100c	4.23	−27	−65	45
Bilateral superior frontal gyrus (BA9)a	346b	4.29	−3	40	43
Bilateral superior frontal gyrus (BA 10)a	204	3.73	−4	54	15
L posterior cingulate	139c	4.54	−19	−62	23
L parahippocampus	46	4.41	−29	−40	−1
R parahippocampus	85	4.16	28	−38	9
R medial temporal cortex	45	3.39	42	−64	0
Bilateral occipital cortex (BA 18,19)a	767	5.89	2	−69	−14
	71	3.62	22	−81	16
	63	3.56	−21	−87	4
R posterior superior temporal sulcusa	362	4.28	47	−40	10
L posterior superior temporal sulcus	177	4.12	−47	−49	7
	45	3.33	−52	−46	−18
R anterior superior temporal lobea	178	4.96	44	−2	−12
L anterior superior temporal lobea	121	4.51	−45	12	−7

Notes: Two of the significant clusters from the analysis extended over multiple brain structures. These larger clusters were parsed into different brain structures in accordance with the oculomotor regions of interest map used in Keedy et al. ³¹ Footnotes indicate which of these 2 large clusters a reported region was part of, if any. Negative values are left, posterior, and inferior to anterior commissure for X, Y, and Z axes, respectively.

^aRegion where reduced activation was found for patients’ baseline activation when restricted to only those with follow-up scans (see supplementary table 4).

^bPortion of a large cluster that totaled 458 voxels.

^cPortion of a large cluster that totaled 322 voxels.

Discussion

Key findings from the present study include pretreatment neural alterations in first episode schizophrenia, and potentially beneficial as well as adverse antipsychotic treatment effects on brain function. First, unmedicated first-episode schizophrenia patients showed reduced activation in the dorsal neocortical attentional control network, whether behavior was elicited by external stimuli, as in the prosaccade task, or guided by internal representations, as in the predictive saccade task. Second, after acute treatment with antipsychotic medication, activation abnormalities diminished in this dorsal frontoparietal attention system. This points to a potential mechanism of treatment efficacy at the functional neural systems level. Third, higher risperidone dose was associated with greater caudate activation after treatment, and caudate activation increases from baseline activation levels showed a trend relationship with improved predictive saccade task performance, suggesting that antipsychotic treatment may enhance the contribution of striatal systems for cognitive control of saccades. Fourth, there was reduced activation within the patient group in dorsolateral prefrontal cortex (DLPFC) after treatment, a finding in line with prior behavioral and imaging work in patients and animal models suggesting an adverse effect of antipsychotic treatment on prefrontal brain systems.^10-12,33-35

Pretreatment Deficits

Before treatment, schizophrenia patients had reduced activation during both saccade tasks in key components of the dorsal neocortical visual attention and saccade control system, including FEF, SEF, and IPS. In prosaccade tasks involving overt attentional processes, these regions are normally robustly activated.^15,17,21 Deficient activation in this brain system is the most consistently observed finding in unmedicated first-episode patients performing saccade tasks.^10,31 This may reflect reduced corticofugal regulation and explain how, under controlled laboratory conditions, unmedicated first-episode patients have speeded visually guided saccade latencies.^19,20 Behavioral deficits were not observed for prosaccades in the present study perhaps due to the small saccade amplitudes and fixed intertrial intervals, whereas larger amplitude saccades and varied trial intervals characterize laboratory studies where such effects are reported.

Additional areas of reduced activation in both tasks included dorsomedial prefrontal cortex (BA 9 and 10) and preSEF. These frontal regions are known to support higher-order cognitive control, including behavioral planning, salience tuning and task switching.^36–38 PreSEF in particular is activated by cognitively demanding eye movement paradigms^39–41 and its alteration pretreatment has been noted in resting state studies of untreated first-episode schizophrenia.⁴²

Impaired performance on the motor learning test, reflected in reduced proportion of predictive saccades pretreatment, could be related to the dorsal neocortical activation deficits, because striatal motor systems were not impaired. Reduced activation in parahippocampus and anterior temporal lobe before treatment also may have impacted task performance via their roles in spatial learning and memory. This pretreatment behavioral abnormality is a novel observation that differs from those of prior laboratory studies, potentially due to the inclusion of several predictive saccade task blocks in the scanner compared with laboratory studies which use a single block, as well as the interleaving and uncued switching between prosaccade and predictive saccade blocks. This switching element in the present study may also explain discrepancies with our prior fMRI study of only prosaccades,¹⁰ where we reported increased activation pretreatment in parietal, occipital, and ventral frontal cortex. We had interpreted increases in the older study as possibly reflecting hyper-sensory responsiveness. In the present study, under-activation prior to treatment could be due to effects of higher task demands (eg, the switching element as well as a higher ratio of active task performance relative to fixation might have differentially increased activation in controls). Further work is needed to clarify these possibilities.

Reduced activation was also observed for unmedicated schizophrenia patients during both prosaccade and predictive saccades in early (BA17-19) and higher-order visual processing regions (posterior superior temporal sulcus and V5). Such effects could be secondary to reduced dorsal attention system engagement, because visual perceptual system activation is enhanced by attentional factors.^43,44 Similar visual system activation deficits have been reported in other first-episode imaging studies.^10,31,45,46

Posttreatment Effects

After treatment, schizophrenia patients had increases in SEF, IPS and occipital cortex activation during the prosaccade task, and in IPS and occipital cortex during the predictive saccade task, all of which were reduced prior to treatment. Thus, effects were in the direction of normalization. By contrast, practice effects in healthy controls exclusively involved reductions in dorsal and medial prefrontal regions. When patient posttreatment activation was contrasted directly with healthy controls’ second scan, no activation deficits were noted for prosaccades, consistent with prior imaging during prosaccades in chronic, treated patients.⁴⁷ Reduction was seen only in temporal cortex for predictive saccades. Taken together, these observations suggest that a significant beneficial effect of antipsychotic treatment may be improvement in the dorsal cortical attention control network. Improved attention may contribute to the greater activation in visual systems after treatment via improved tuning of sensory system responsivity to task-relevant stimuli.

Another posttreatment observation was that higher risperidone doses were associated with greater caudate nucleus activation in the prosaccade task, and this was also found at a trend level for the predictive saccade task. While the caudate did not show significant activation change with treatment during saccade tasks in the group analysis, the observation of greater activation for patients on higher doses of risperidone is of interest given the robust binding and neurophysiological effects of antipsychotic medication in the striatum. Several studies of resting brain function have shown enhanced activity in striatum after antipsychotic treatment.^42,48,49 One additional observation related to the caudate nucleus was that during predictive saccades, those patients with greater behavioral improvement had a trend toward greater increases in right caudate activation relative to pretreatment, consistent with the view that treatment-related changes in caudate activation may be important for supporting improved predictive learning after treatment.

The group-level improvement in predictive saccade rates pretreatment to posttreatment, with patient-control differences no longer being significant, is a potential cognitive benefit of treatment seen at the behavioral level. Prior laboratory studies of predictive saccade tasks with medicated, chronically ill schizophrenia patients report that such patients make significantly faster or more frequent anticipatory saccades relative to controls.^50-52 Given this, it is possible we have captured an intermediate stage of a treatment effect (our ~6 wk of treatment vs years of treatment in chronic patient samples) on predictive saccade task performance.

We also observed reduced DLPFC activation after antipsychotic treatment, consistent with our previous fMRI study using a saccade task before and after treatment.¹⁰ This was the only area with reduced activation in patients posttreatment, and no change in DLPFC was found for healthy participants at follow-up. The reduction in DLPFC activation posttreatment for patients was associated at a trend level with performance, such that those with greater reduction in DLPFC activation after treatment showed the least performance improvement. Altogether, the pattern of findings suggests that beneficial treatment effects in striatum and in dorsal cortical attention systems occur, but at the same time there are adverse effects of treatment in DLPFC evident in the task conditions employed in the present study. This pattern of findings might be interpreted in the context of the expanded dopamine hypothesis of schizophrenia, which postulates hyperdopaminergia in striatum and hypodopaminergia in frontal cortex.⁵³ Such deficits might influence response to a D2 antagonist such that striatal function may be relatively normalized but frontal cortex may shift further into a state of hypofunction via local effects or changes in thalamic drive. Treatment-related DLPFC reduction is consistent with prior studies showing reduced spatial accuracy of predictive saccades after treatment in first-episode schizophrenia patients¹² and in chronically treated patients.^50,54,55 Measurement of saccade accuracy in the scanner was not sufficient to test for changes in the small saccade amplitudes elicited by our tasks to test this hypothesis in the present study. Other lines of work are consistent with reduced prefrontal function after antipsychotic treatment and its impact on cognitive performance. Longitudinal laboratory studies of first-episode patients performing a spatial working memory task show increased impairment after treatment,^11,56 paralleling similar effects of antipsychotics in nonhuman primates,⁵⁷ and schizophrenia patients’ Wisconsin Card Sort performance has been shown to be negatively correlated with antipsychotic dose.³³

Limitations of this study include the relatively small sample. We may have lacked statistical power to detect some abnormalities pretreatment or some adverse or beneficial drug effects. This might be particularly relevant for areas with reduced activation pretreatment where more modest improvement might not have been detected. Another limitation was clinician-judged dosing, so that clinical features and drug dose are, to a degree, confounded. Also, posttreatment effects cannot be definitively attributed to drug vs state of illness without placebo control. The clinical reality is that the patients require active treatment, yet it remains a limitation for data interpretation. Finally, attention and learning are complex constructs requiring multiple testing approaches to fully understand their alterations in psychosis and in relation to treatment, so further work is needed to replicate and extend our observations to other areas of neurocognitive processing.

Supplementary Material

Supplementary material is available at http://schizophreniabulletin.oxfordjournals.org.

Funding

National Institute of Mental Health (R01MH62134, K23MH083126, K23MH092702, K08MH083888); Brain and Behavior Research Foundation (Young Investigator Awards to S.K.K, J.L.R.).