INTRODUCTION
The term “executive function” is used to cover a variety of fractionated cognitive processes concerned with the control, organization and sequencing of higher cognition. This umbrella term includes such functions as working memory, initiation, response inhibition, problem solving, and planning (Burgess & Simons, Reference Burgess, Simons, Halligan and Wade2005). Studies have estimated that 27–46% of people with schizophrenia have selective “executive” profiles and 54–90% have at least one executive impairment (Chan et al., Reference Chan, Chen, Cheung, Chen and Cheung2006a,Reference Chan, Chen and Lawb; Johnson-Selfridge & Zalewski, Reference Johnson-Selfridge and Zalewski2001; Kremen et al., Reference Kremen, Seidman, Faraone, Toomey and Tsuang2004). Executive dysfunction is associated with poor social outcome (Kopelowicz et al., Reference Kopelowicz, Liberman, Ventura, Zarate and Mintz2005; Laes & Sponheim, Reference Laes and Sponheim2006) and has become the target for neurorehabilitation (Wykes et al., Reference Wykes, Reeder, Landau, Everitt, Knapp, Patel and Romeo2007a). The impairment has been related neurobiologically to the connectivity between the prefrontal cortex and the striatal circuitry, forming the substrate for organization processes that control cognition (Elliott, Reference Elliott2003).
Several factors may impact on the characteristics of executive dysfunction in schizophrenia. A main consideration is the relationship with symptoms, linked both theoretically and empirically with the pattern of neuropsychological impairment (Hill et al., Reference Hill, Ragland, Gur and Gur2002; Kremen et al., Reference Kremen, Seidman, Faraone, Toomey and Tsuang2004; Liddle, Reference Liddle1987a,Reference Liddleb). Models of schizophrenia have suggested a distinct syndromic structure (Emsley et al., Reference Emsley, Rabinowitz and Torreman2003; Liddle, Reference Liddle1987a,Reference Liddleb; Mass et al., Reference Mass, Schoemig, Hitschfeld, Wall and Haasen2000) such as a demarcation into three syndromes, namely reality distortion, disorganization, and psychomotor poverty (negative symptoms). The latter two of these are associated with executive dysfunction (See Donohoe & Robertson, Reference Donohoe and Robertson2003), as in the theory by Frith (Reference Frith1992): Disorganization symptoms arise from impaired inhibition of habitual responses when plans must be constructed and implemented using working memory, whereas psychomotor poverty results from deficits in the initiation of activities due to impaired initiation of plans. In studies of single symptoms, both syndromes have been associated with impaired verbal initiation and working memory and disorganization also with attention, inhibition, discourse planning and monitoring (Hoffman et al., Reference Hoffman, Stopek and Andreassen1986; Liddle & Morris, Reference Liddle and Morris1991; Pantelis et al., Reference Pantelis, Stuart, Nelson, Robbins and Barnes2001). Although the symptom structures have been debated, all recent refinements of symptom models retain the distinction between disorganization and psychomotor poverty. Previous studies have compared profiles across domains in relation to symptoms, or have provided fine-grained analyses within executive functions (Chan et al., Reference Chan, Chen, Cheung, Chen and Cheung2006a,Reference Chan, Chen and Lawb; Hill et al., Reference Hill, Ragland, Gur and Gur2002; Hutton et al., Reference Hutton, Puri, Duncan, Robbins, Barnes and Joyce1998; Kremen et al., Reference Kremen, Seidman, Faraone, Toomey and Tsuang2004; Liddle & Morris, Reference Liddle and Morris1991), but there has been no direct comparison of profiles on a broad range of executive processes in relation to both disorganization and psychomotor poverty. Due to the over-reliance on a “unitary view of executive function in schizophrenia”, theoretical accounts that dissociate executive processes, such as Frith's theory, have remained largely unexplored (Donohoe & Robertson, Reference Donohoe and Robertson2003).
A second consideration is chronicity, which cuts across symptom distinctions. To study this issue, researchers have compared functioning in first episode and chronic schizophrenia across different domains. First episode schizophrenia shows executive dysfunction at this early stage, with some degree of clinical heterogeneity (Chan et al., Reference Chan, Chen and Law2006b; Joyce et al., Reference Joyce, Hutton, Mutsatsa and Barnes2005; Joyce & Roiser, Reference Joyce and Roiser2007), but less impairment than is found in chronic schizophrenia (Chan et al., Reference Chan, Chen, Cheung, Chen and Cheung2006a; Saykin et al., Reference Saykin, Shtasel, Gur, Kester, Mozley, Stafiniak and Gur1994). Profiles have varied between studies, with parallel flat profiles of diffuse general impairment, parallel nonflat profiles with selective impairments, and selective impairments specific to chronic schizophrenia (Albus et al., Reference Albus, Hubmann, Ehrenberg, Forcht, Mohr, Sobizack, Walheim and Hecht1996; Blanchard & Neale, Reference Blanchard and Neale1994; Chan et al., Reference Chan, Chen, Cheung, Chen and Cheung2006a,Reference Chan, Chen and Lawb; Saykin et al., Reference Saykin, Shtasel, Gur, Kester, Mozley, Stafiniak and Gur1994). These variations might result from studies that collapse test scores across broad domains. This is problematic for executive functions that fractionate into uncorrelated processes (Burgess & Simons, Reference Burgess, Simons, Halligan and Wade2005). Nevertheless, the main pattern appears to be that profiles are broadly the same in the two stages of the illness, suggesting a core pattern of dysfunction.
Executive profiles are less clearly linked than memory and globally impaired profiles to symptom subgroups, perhaps due to the unitary approach, but are associated with cognitive and symptom severity and illness duration (Chan et al., Reference Chan, Chen, Cheung, Chen and Cheung2006a,Reference Chan, Chen and Lawb; Hill et al., Reference Hill, Ragland, Gur and Gur2002; Kremen et al., Reference Kremen, Seidman, Faraone, Toomey and Tsuang2004). Categorization of subgroups by neuropsychological profile in all cases confounds the relationship between symptoms and chronicity. Hence the current study explores the severity and profile of executive functioning in relation to disorganization and psychomotor poverty. Simultaneously, it investigates the distinction between first episode and chronic schizophrenia. An overlapping sampling method is used in which a general pool of people with schizophrenia are split either by symptoms or according to chronicity, hence the simultaneous comparisons. Furthermore, the study determines whether symptom groupings differentiate on the basis of a fine-grained analysis of executive function and whether the early and late profiles are similar. Four main aspects of executive functioning are explored: working memory, planning and strategy formation, response initiation, and response inhibition. Based on theoretical accounts of schizophrenia that consider fractionated executive function (e.g., Frith, Reference Frith1992), disorganization was predicted to be associated with a broader range of executive deficit whilst psychomotor poverty would be associated with impairments in working memory and response initiation. The study explores whether the overall pattern of impairment remains when comparing first episode versus chronic patients.
METHOD
Design
Executive functioning was compared in a cross-sectional sample of people with schizophrenia and healthy controls, the former split by symptoms or chronicity for two parallel analyses. The healthy control group provided a comparison for group effects and a normative reference for profile analyses. General intellectual functioning was controlled.
Participants
All participants were sampled from the same inner-city geographical area and were aged 18–65, with a screen IQ estimate over 70, using the National Adult Reading Test-Restandardized (Nelson & Willison, Reference Nelson and Willison1991), English as a first language, no neurological problems or head injury, and no current substance abuse. The 28 control participants were screened by self-report on past mental health problems and contact with psychiatric services and were recruited provided they had no psychiatric history.
The schizophrenia participants were hospitalized or were outpatients and satisfied DSM-IV criteria confirmed using: DSM-IV checklist, clinical notes and consultant psychiatrist collateral information. People with schizoaffective disorder were excluded. Symptoms were assessed using SAPS and SANS (Andreassen, Reference Andreassen1984a,Reference Andreassenb). Disorganization was defined as scoring at least 8 from SAPS formal thought disorder and inappropriate affect and SANS poverty of content of speech items. Psychomotor Poverty was defined as having at least one core symptom of alogia or flat affect (summed score at least 6) and one other negative symptom (summed score at least 3 on any remaining SANS subscale) (Liddle, Reference Liddle1987a). First episode participants comprised all eligible participants with schizophrenia recruited as part of a large population-based study of first-episode psychosis (Dazzan et al., Reference Dazzan, Morgan, Orr, Hutchinson, Chitnis, Suckling, Fearon, Salvo, McGuire, Mallett, Jones, Leff and Murray2004). DSM-IV diagnoses were additionally confirmed through consensus and 6-month follow-up. Participants were identified within 1 week of their first psychiatric contact for psychosis and were assessed as soon as their symptoms had stabilized (see Table 1). Chronic illness was defined as at least 5 years from first psychiatric contact for psychosis, beyond any period of initial deterioration. Chronicity was verified by self-report and the supporting collateral clinical information.
Table 1. Sociodemographic and clinical characteristics for the symptom sample
Note
Mean (Standard deviation) unless otherwise stated. Medication dose was converted to a standard total percentage of the maximum recommended dose (Maudsley and Bethlem Prescribing Guidelines 1996). Parental SES was divided into 3 categories 1 = professional/managerial, 2 = skilled, 3 = semi-skilled/unskilled. * Different due to group definition.
The study was reviewed and ethical committee approval was obtained. After a description of the study, all the participants gave written informed consent.
A total of 71 participants were allocated to either or both samples according to fulfillment of membership criteria. The chronicity sample comprised 22 first episode and 35 chronic participants, while the symptom sample encompassed 29 designated as disorganization and 29 as psychomotor poverty. Some participants fulfilled the criteria for only one sample because disorganization/psychomotor poverty symptoms were absent or too mild to fulfill symptom criteria (n = 13) or because illness length lay between first episode and chronic sample criteria (n = 14).
Measures
Socioeconomic status
Parental socioeconomic status was derived from self-report of paternal occupation at participant's birth, which was categorized into one of five broadly defined categories from 1 (professional) to 5 (unskilled) (Office of Population Consensus' and Surveys, 1991).
General intellectual function
Intellectual function was assessed with the NART-Restandardized and current IQ with a recognized WAIS-R short form (Canavan et al., Reference Canavan, Dunn and McMillan1986).
Executive function
Tests were selected to measure a range of well defined primary executive processes known to be affected in schizophrenia, including working memory, planning/strategy use, response initiation, and response inhibition, measured across verbal and spatial domains (Baddeley, Reference Baddeley, Tulving and Craik2000).
Working memory
The tasks were chosen to distinguish between verbal and spatial maintenance and manipulation processes (Baddeley, Reference Baddeley, Tulving and Craik2000; D'Esposito et al., Reference D'Esposito, Postle, Ballard and Lease1999).
For verbal working memory the digit span forward test was used to measure maintenance (Baddeley, Reference Baddeley, Tulving and Craik2000), with span defined as the longest sequence recalled correctly on 2/3 trials. Manipulation was measured using the Verbal Working Memory Word Span task, a simplified version of the Daneman and Carpenter (Reference Daneman and Carpenter1980) sentence span task. The task requires remembering words, with interleaving decision making concerning recognition of previously presented words and relies on maintenance with additional manipulation requirements (Greenwood, Reference Greenwood2000). Single word lists were presented consecutively on a computer screen, starting with two words and increasing by one word every four sequences. After list presentation, recognition memory for the probe word was tested with selection of a “yes” or a “no” box on the screen. The probe was a target word on 50% of occasions, a word closely associated with the target on 25% and a distracter on 25% of occasions. Immediately after selection, the participants were required to recall the word list. The manipulation measure was the longest sequence of words (word span) recalled correctly on at least 3 of 4 trials.
For Spatial Working Memory, the Executive Golf Task, a version of the Spatial Working Memory task (SWT) (Hutton et al., Reference Hutton, Puri, Duncan, Robbins, Barnes and Joyce1998; Joyce et al., Reference Joyce, Hutton, Mutsatsa and Barnes2005; Owen et al., Reference Owen, Morris, Sahakian, Polkey and Robbins1996) was used that encompasses the distinction between maintenance and manipulation. This is a naturalistic version of the task with three dimensional graphics (Feigenbaum et al., Reference Feigenbaum, Polkey and Morris1996; Miotto et al., Reference Miotto, Bullock, Polkey and Morris1996), which has been used in schizophrenia (Kravariti et al., Reference Kravariti, Morris, Rabe-Hesketh, Murray and Frangou2007; Toulopoulou et al., Reference Toulopoulou, Morris, Rabe-Hesketh and Murray2003, Reference Toulopoulou, Grech, Schulze, McDonald, Morris, Rabe-Hesketh and Murray2004). This task (see previous papers for description) involves a series of searches for hidden targets on a computer screen, with nonreturn to the successful target on subsequent searches. A maintenance measure is derived from errors in returning to the same location during a single search, whilst a measure of manipulation is based on errors that occur when the participant returns to a previously successful location between searches. Studies have shown that within search errors dissociate from between search errors in patients with focal temporal and frontal lesions (Feigenbaum et al., Reference Feigenbaum, Polkey and Morris1996; Miotto et al., Reference Miotto, Bullock, Polkey and Morris1996).
Planning
This was measured with the 3D-Computerised Tower of London task (Kravariti et al., Reference Kravariti, Morris, Rabe-Hesketh, Murray and Frangou2003, Reference Kravariti, Morris, Rabe-Hesketh, Murray and Frangou2007; Morris et al., Reference Morris, Rushe, Woodruffe and Murray1995), which is based on the Shallice Tower of London task (Reference Shallice1982) and is similar to the Stockings of Cambridge computerized task (CANTAB: Robbins et al., Reference Robbins, James, Owen, Sahakian, McInnes and Rabbitt1994). The task involves moving discs around a display in a certain sequence, using a touch sensitive screen, to achieve a target arrangement. The main score used in this study is the Mean Moves above Minimum required to solve 4 or 5 move problems. This measure, used in previous studies with similar procedures is sensitive to frontal lobe damage affecting planning ability (e.g., Owen et al., Reference Owen, Downes, Sahakian, Polkey and Robbins1990).
Strategy use
Verbal strategy formation was assessed using a modified version of the Hayling task (Burgess & Shallice, Reference Burgess and Shallice1996). The original task required the completion of sentences with a single meaningless completion word and was modified to include sentences with both high and low dominance prepotent responses. Common strategies are to look for items in the room (visual) or to link successive responses semantically (verbal). A total strategy scoring method is based on the procedure by Burgess and Shallice (Reference Burgess and Shallice1996). Spatial strategy use was assessed using a measure of strategy formation on the Executive Golf Task. More efficient search paths involved sampling proximal locations in a sequential strategic manner. A higher score constituted poorer strategy use. This measure shows selective strategy formation deficits in patients with frontal lobe lesions (Owen et al., Reference Owen, Morris, Sahakian, Polkey and Robbins1996) and dissociates strategy formation and working memory deficits in schizophrenia (e.g., Pantelis et al., Reference Pantelis, Stuart, Nelson, Robbins and Barnes2001).
Response initiation
This was measured using a modified verbal fluency task which alternated between searches by letter (FAS version) and by category (Animals, Fruits, Body Parts) with 30-second trials for each (Greenwood, Reference Greenwood2000; Spreen & Benton, Reference Spreen and Benton1977).
Response inhibition
This was estimated verbally by taking the number of incorrect meaningful completion words produced on the modified Hayling task. This response inhibition estimate has been shown through factor analysis to be distinct from the estimate of strategy use derived using the same test (Burgess & Shallice, Reference Burgess and Shallice1996). Spatially, response inhibition was assessed using the incompatible minus the compatible response latency on the Complex Reaction Time (4-button) task, which has been used previously as a measure of response inhibition (Wykes et al., Reference Wykes, Sturt and Katz1990).
Medication dose
To obtain a single medication dose across typical and atypical neuroleptics, dose was converted to a standard total percentage of the maximum recommended dose using the Maudsley and Bethlem Prescribing Guidelines (1996).
Statistical Analyses
Assessment of group differences in level of function
Group differences in executive functioning were investigated using MANCOVAs of the executive variables. WAIS-R IQ was covaried to control for the relationship between schizophrenia and general cognitive impairment, and for the relationship between tests due to cognitive congruence. This increased the specificity to explore the relationship between schizophrenia and distinct executive functions. The analysis allowed for IQ to have a different effect on each of the measures under investigation. The assumed linear relationship between executive functions and IQ within each group was assessed using residual plots and no evidence of nonlinearity was found. The linear effect of IQ did not vary between groups in the respective interaction tests, which were not statistically significant at the 5% level. Age and education were not controlled as the former covaries with chronicity and the latter is disrupted as a side effect of the onset of schizophrenia. Post hoc pairwise group comparisons were carried out on group differences with multistage Bonferroni corrections to control for multiple group comparisons.
Assessment of differences within executive profiles
Executive functioning profiles were investigated by transforming executive variables to Z-scores by reference to the normal controls and then analyzing the effect of the within-subject factor, executive process, and its interaction with the between-subject factor, group. Profile analyses were conducted on the schizophrenia groups using generalized estimating equations (GEE) to fit linear-models for the Z-scores that assumed normally distributed errors with constant variances (Hardin, Reference Hardin, Everitt and Howell2005). The model allowed for correlations between the repeated Z-scores by assuming an exchangeable covariance structure. To account for a more complex covariance structure and nonconstant errors, the robust sandwich estimator was used (Williams, Reference Williams2000).
For each profile analysis, the interaction between group and executive process was investigated. If this was significant it was maintained in the model and further investigations were carried out for each group separately. Otherwise the main effect of executive process was assessed across groups. An additional interaction term between executive process and IQ was fitted.
Assessment of specific islets of strength or deficit
Four subdomains were derived to reflect those identified by Shallice and Burgess (Reference Shallice and Burgess1996). Planning and strategy use were considered part of the New Schema Construction subdomain. Response inhibition was incorporated within the Assessment and Verification of Schema subdomain. The third subdomain, Implementation of Schema within Working Memory, was considered to incorporate both working memory and response initiation which are hypothesized to be differentially associated with different symptoms and so were separated. Each domain score (planning/strategy use, response inhibition, working memory, response initiation) was compared with the average of all other process scores while holding IQ constant at 100. When the GEE analysis identified an interaction between group and executive process, the process comparisons were conducted within each group separately, otherwise comparisons were carried out across groups. The multi-stage Bonferroni procedure was used to control for multiple domain or group by domain comparisons.
Assessment of the impact of symptoms on the chronicity analysis
As was expected due to sampling factors in our cross-sectional design, symptoms were associated with chronicity. Specifically, disorganization symptoms were greater in the chronic group (see Table 2). As this might affect chronicity profiles, a GEE analysis of the chronicity sample controlled for more severe forms of disorganization using a new severe thought disorder factor which was entered into the analysis. The criterion for severe thought disorder was a score of 4 or more on at least one of the SAPS thought disorder measures. Ten participants with chronic and two with first episode schizophrenia fulfilled criteria for high thought disorder severity. No control was made for psychomotor poverty which did not differ between chronicity groups.
Table 2. Sociodemographic and clinical characteristics for the chronicity study
Note
Medication dose was converted to a standard total percentage of the maximum recommended dose (Maudsley and Bethlem Prescribing Guidelines 1996). Parental SES was grouped into 3 categories 1 = professional/managerial, 2 = skilled, 3 = semi-skilled/unskilled. * Different due to group definition.
Assessment of the impact of medication on executive function
GEE analyses of medication dose [high (greater than 50% of the maximum dose) compared with low dose] and group (standard neuroleptics, clozapine, and other atypicals) assessed medication effects. In each case, standardized scores were derived using one group (Low dose or Standard Neuroleptics respectively) as the reference.
All analyses were conducted using STATA (Statistical Data Analysis) version 7.0.
RESULTS
Demographics
The groups were balanced for sex and parental socioeconomic status, which is purportedly the optimal demographic matching variable for schizophrenia (Keefe, Reference Keefe1995; see Tables 1 and 2). The first episode group was, by design, younger than the chronic sample, but the difference is in an age range where substantive age-related cognitive differences are not expected. All schizophrenia groups had significantly fewer years of education and lower IQ than controls. The IQ of the control sample did not differ significantly from their demographically predicted premorbid IQ. This IQ estimate is largely determined by social class which was matched between the schizophrenia and control samples (Crawford & Allan, Reference Crawford and Allan1997). The control sample, therefore, was estimated to perform at the level that the schizophrenia sample might otherwise have attained, based on demographics, had they not developed schizophrenia. There were no education or IQ differences between any of the schizophrenia groups. The schizophrenia groups were well balanced on clinical characteristics, except for the expected symptom differences. The chronic group demonstrated statistically greater disorganization and general negative symptoms (asociality, anhedonia, avolition, and apathy) compared with first episode schizophrenia. They were also well-balanced for medication use with no statistically significant differences between groups in neuroleptic dose or the proportion of participants who were taking atypical neuroleptics.
Symptoms Sample
Group differences in level of function
Figure 1 shows the executive profiles. A significant group effect was detected after controlling for IQ [Wilks lambda = 0.36, F(20,128) = 4.21; p < .001]. Post hoc pairwise between group comparisons revealed poorer performance in each schizophrenia group compared with controls [Wilks lambda = 0.32, F(10,39) = 8.13; p < .001 for psychomotor poverty and Wilks lambda = 0.55, F(10,41) = 3.29; p = .003 for disorganization] and a trend for poorer performance in the disorganization compared with the psychomotor poverty syndrome [Wilks lambda = 0.67, F(10,38) = 1.89; p = .07 compared with α = .05/1].
Fig. 1. Executive Impairment Profiles in the Psychomotor Poverty (n = 29) and Disorganization syndromes (n = 29) of schizophrenia compared with normal control performance (n = 28) and with current IQ held constant at 100. Working Memory [ds = digit span, swe = spatial within search errors, vwm = verbal working memory, sbe = spatial between search errors]; Planning and Strategy [pl = planning, vss = verbal strategy score, sss = spatial strategy score]; Response Initiation-Inhibition [ri = response initiation, vinhib = verbal inhibition, sinhib = spatial inhibition].
Differences within executive profiles
The profile comparisons revealed a significant interaction between executive process and group [χ2(9) = 25.01; p = .003] after adjusting for the interaction between executive process and IQ [χ2(9) = 24.85; p = .003]. Post hoc and selectivity analyses revealed significant within-group effects of executive process in both the psychomotor poverty and the disorganization groups [χ2(9) = 96.9; p < .0001; χ2(9) = 55.6; p < .0001] consistent with different executive profiles in these groups.
Assessment of specific islets of strength or deficit
The analyses revealed a selective responds initiation impairment within both the psychomotor poverty and disorganization groups compared with the average of all other processes [(estimated difference in mean Z-scores = −1.61; Z = −8.14; p < .001; CI(95) = (−1.99 to −0.22) for Psychomotor Poverty and = −0.79; Z = −2.67; p = .007, CI(95) = (−1.36 to −0.21) for Disorganization) compared, according to the multistage Bonferroni procedure, with α = .05/7 and .05/6, respectively]. Within the Psychomotor Poverty group, planning and strategy use was relatively intact compared with the average of all other processes [estimated difference in mean Z-scores = 0.50; Z = 3.51; p < .0001; CI(95) = (0.22–0.77) compared with α = .05/8]. However, Figure 1 suggests that this relative strength was determined by planning and spatial strategy use.
Chronicity Comparison
Group differences in level of function
Lower scores were observed in the chronic compared with the first episode group on all but one measure (Figure 2). A significant group effect was found after controlling for IQ [Wilks lambda = 0.50; F(20,130) = 2.5; p < .001]. Post hoc pair-wise between-group comparisons revealed poorer performance in chronic schizophrenia [Wilks lambda = 0.48; F(10,47) = 5.01; p < .001] and a trend for poorer performance in first episode schizophrenia compared with controls [Wilks lambda = 0.59; F(10,34) = 2.39; p = .029 compared with α = .05/2] but no statistical difference between chronicity groups [Wilks lambda = 0.71; F(10,39) = 1.59; p = .15]. As there was no group difference between first episode and chronic schizophrenia, the impact of age on the group main effect was not investigated further.
Fig. 2. Executive Impairment Profiles in First episode (n = 22) and Chronic Schizophrenia (n = 35) compared with normal control performance (n = 28) and with current IQ held constant at 100 (see Figure 1 for measure labels).
Differences within executive profiles
There was no interaction between group and executive process [χ2(9) = 14.5; p = .106], but there was a significant effect of executive process [χ2(9) = 55.7; p < .0001] and a significant interaction between executive process and IQ [χ2(9) = 35.9; p < .0001], consistent with parallel nonflat executive profiles in first episode and chronic schizophrenia. The two chronicity groups were, therefore, combined and the remaining post hoc and selectivity analyses were conducted within-groups.
Assessment of specific islets of strength or deficit
The analyses revealed a selective impairment in response initiation compared with each other process [(estimated difference in mean Z-scores = −1.03; Z = −5.62; p < .001, CI(95) = (−1.39 to −0.67) compared, according to the multistage Bonferroni procedure, to α = .05/4)].
Assessment of the impact of controlling for symptoms on the chronicity analysis
Disorganization symptoms but not psychomotor poverty symptoms differed between the two chronicity groups. It was predicted that the executive profiles would, therefore, be more similar in first episode and chronic schizophrenia if the distribution of disorganization symptoms was controlled. The results for the level of impairment analysis remained unchanged with a significant group effect [Wilks lambda = 0.53; F(20,128) = 2.40; p = .002], a post hoc between-group difference in chronic schizophrenia [Wilks lambda = 0.50; F(10,46) = 4.6; p < .001] and a trend for a difference in first episode schizophrenia compared with healthy controls [Wilks lambda = 0.55; F(10,33) = 2.25; p = .039 compared with α = .05/2] but no difference between chronicity groups [Wilks lambda = 0.73; F(10,38) = 1.42; p = .21].
The profile analysis also remained unchanged with a significant effect of executive process [χ2(9) = 51.2; p < .0001] so that even after adjusting for disorganization symptoms the within-group profiles were significantly nonflat.
The analysis of impairments, after controlling for disorganization, again indicated a significant response initiation impairment [estimated difference between domain and average of remaining domains = −1.08; Z = −5.53; p < .001; CI(95) = (−1.47 to −0.70)] but now also indicated relatively intact working memory (estimated difference = 0.32; Z = 2.45; p = .014 compared with α = .05/3; CI(95) = (.065–0.58). A comparison between Figures 2 and 3 revealed that this reflected the more similar profiles in first episode and chronic schizophrenia, particularly in working memory, after controlling for disorganization.
Fig. 3. Executive Impairment Profiles in First Episode Schizophrenia and Chronic Schizophrenia with lower disorganization severity compared with normal control performance and with current IQ held constant at 100. See Figure 1 for measure labels.
Assessment of the impact of medication on executive function
There were no significant main effects of either medication dose or group on executive function (p's > .10).
DISCUSSION
The study demonstrated fractionated profiles of executive impairments, after controlling for IQ, which differed in relation to symptom type but not in relation to chronicity. The differential pattern of executive impairments in schizophrenia is thus determined by symptoms and not chronicity. Executive impairments in the disorganization group were generally greater and encompassed a broader range of processes than in the psychomotor poverty group. In the disorganization group, planning/strategy use was as impaired as the majority of other processes while in the psychomotor poverty group it was more intact than other processes. When disorganization was controlled, profiles were closer and working memory was statistically more intact across first episode and chronic schizophrenia, which suggests an association between disorganization and working memory impairments. Response initiation was selectively impaired to different degrees in all schizophrenia groups but this impairment was greatest in the psychomotor poverty group. The results support the original theoretical proposition by Frith (Reference Frith1992) that negative symptoms may be underpinned by impaired initiation of novel responses while disorganization symptoms incorporate a broader range of impairments including inhibition and planning. This pattern of multiple selective deficits within executive function is consistent with predicted differential fractionation of executive function in different symptom groups (Chan et al., Reference Chan, Chen, Cheung, Chen and Cheung2006a) and with patterns described previously across different cognitive domains in schizophrenia as a whole (Sullivan et al., Reference Sullivan, Shear, Zipursky, Sagar and Pfefferbaum1994).
In contrast, consistent with some but not all studies, profiles were parallel and did not differ statistically in severity between first episode and chronic schizophrenia. Most studies focused on cognition more broadly rather than on executive processes (Censits et al., Reference Censits, Ragland, Gur and Gur1997; Saykin et al., Reference Saykin, Shtasel, Gur, Kester, Mozley, Stafiniak and Gur1994). Some studies showed significantly greater executive impairment in chronic compared with first episode schizophrenia (Albus et al., Reference Albus, Hubmann, Ehrenberg, Forcht, Mohr, Sobizack, Walheim and Hecht1996). Others showed a wide range of executive impairments at first episode but a disproportionate number (46%) with intact executive function (Chan et al., Reference Chan, Chen and Law2006b). Here, the chronic group performed more poorly than the first episode sample on 9 of 10 measures so the lack of statistical significance may have reflected some lack of power.
This research is novel in demonstrating in a single study that schizophrenia as a whole, whether first episode or chronic, appears to be characterized by a single executive profile. Yet, this executive profile seems likely to reflect the symptom make-up of the samples. The similar impairment pattern in first episode and chronic samples suggests a similar but attenuated proportional mix of different symptom-specific executive impairment profiles in first episode compared with chronic schizophrenia. The less severe impairment profile in first episode schizophrenia may have reflected the intact executive performance in those with first episode schizophrenia who will not become chronic. This attenuated executive impairment in first episode schizophrenia may either result from a proportion of the first episode sample with flat, intact profiles similar to controls or from a characteristic executive profile throughout the sample which nevertheless lies above the control mean and in the intact range for a proportion of the first episode sample. This intact cognition group may be those with good outcomes.
Poor cognition is a recognized marker of poor outcome in schizophrenia from first episode and appears to contribute to this outcome independently of symptoms (Dickinson & Coursey, Reference Dickinson and Coursey2002; Moritz et al., Reference Moritz, Krausz, Gottwalz, Lambert, Perro, Ganzer and Naber2000). There are difficulties in identifying stable symptoms that are predictive of poor outcome at first episode (Fujii & Wylie, Reference Fujii and Wylie2003; Malla et al., Reference Malla, Norman, Takhar, Manchanda, Townsend, Scholten and Haricharan2004) and the predictive power of executive functions on first episode outcome appears short-lived (Goldstein et al., Reference Goldstein, Giovannetti, Schullery, Zuffante, Lieberman, Robinson, Barr and Bilder2002). This may reflect the tendency for executive functions, negative and disorganization symptoms to improve (Schuepbach et al., Reference Schuepbach, Keshavan, Kmiec and Sweeney2002) with improved prospective outcome. The combined stable presence of these executive impairments and symptoms at first episode may together indicate poor outcome. The targeted remediation of these candidate “chronicity” markers may prove beneficial in remediating the poor outcome in these symptom groups (Wykes et al., Reference Wykes, Reeder, Landau, Everitt, Knapp, Patel and Romeo2007a,Reference Wykes, Newton, Landau, Rice, Thompson and Frangoub).
This study has several limitations. A single balanced sample across chronicity and symptoms would have allowed the measurement of their independent effects and interactions on profiles. However, even in the current large sample, few first-episode participants met symptom criteria so the power of the study to detect unadjusted group differences would be compromised. The measures were selected to encompass a comprehensive range of executive processes within four broad domains that are commonly impaired in schizophrenia, but some processes were not tested directly. The individual measures were selected to most closely reflect a single individual process and general cognition was controlled but several processes may still have contributed to performance on each individual measure (Jaeger et al., Reference Jaeger, Czobor and Berns2003). Most schizophrenia participants were taking neuroleptic medication, but there were no differences in dose or medication type between schizophrenia groups. Chronicity did not interact with executive functioning, making it unlikely that life-time medication exposure, which would covary with chronicity, mediated differences in executive functioning. This is consistent with findings that executive dysfunction is present in first episode neuroleptic naïve schizophrenia and persists after controlling for medication use in chronic samples (Chan et al., Reference Chan, Chen, Cheung, Chen and Cheung2006a,Reference Chan, Chen and Lawb).
This study addresses issues identified in previous investigations of executive function in schizophrenia (Chan & Toulopoulou, Reference Chan, Toulopoulou and French2006; Donohoe & Robertson, Reference Donohoe and Robertson2003). The approach to executive function is theory-driven and incorporates the fractionation of executive functions. This specificity of executive process impairment from general impairment is further enhanced through the covariation of current intellectual function and of component executive processes. Symptom subgroups and illness duration were investigated simultaneously to separate their impact on quantitative (severity) and qualitative executive impairments in schizophrenia (Kremen et al., Reference Kremen, Seidman, Faraone, Toomey and Tsuang2004). The study includes a sizeable disorganization group, and the recruitment of a first episode sample early in their first presentation minimizes the impact of long-term medication effects and onset associated cognitive decline.
The fractionation of executive functions precludes a single executive profile as might occur for other functional domains. However, the current limitations in our theoretical and empirical knowledge of executive function in schizophrenia tend to prevent more detailed subtyping by fractionated executive processes. An alternative approach, specifically when considering executive function in schizophrenia, is to consider subgroups defined by symptoms which are then investigated for their differential executive profiles. The refinements in the understanding of symptom subtypes which now go well beyond early positive-negative dichotomies, yet still retain the negative and disorganization distinction, combined with a greater understanding of vulnerability factors linked to symptom subtypes, provides greater empirical support for a symptom approach. These symptom distinctions appear reliable across diagnostic boundaries and in their associations with executive impairments (Kravariti et al., Reference Kravariti, Dixon, Frith, Murray and McGuire2005). The merits of this approach lie in the advanced understanding of the relationship between executive function and the phenomenology of schizophrenia, with predicted advances in understanding and remediating chronic schizophrenia with poor functional outcome.
ACKNOWLEDGMENTS
This study was supported by a PhD studentship from the Medical Research Council. We thank Professors Robin Murray and Julian Leff, Institute of Psychiatry, for enabling collaboration with the AESOP project to recruit people with first episode psychosis and for providing consensus diagnoses in this group.
