Participants

The 78 participants with first-episode psychosis included in the following analyses satisfied ICD–10 ⁶ criteria for schizophrenia (n = 44), schizoaffective disorder (n = 2) and bipolar disorder/mania (n = 32) as determined by clinical consensus reviews of patient histories and structured interviews using the Schedule for Clinical Assessment in Neuropsychiatry (SCAN). ^{Reference Wing, Babor, Brugha, Burke, Cooper and Giel7} The majority (91%) were being prescribed antipsychotics at the time of testing (5% typical antipsychotics). Healthy comparison participants (n = 67) were recruited from general practitioner surgeries and by advertising on the Belfast City Hospital site. Controls were selected to be similar to participants in the schizophrenia and the bipolar disorder groups on distributions of gender and age.

Procedure

Assessments were undertaken within 3 months of the individual's first presentation to psychiatric services (median 38 days). Eighty-six per cent of people were in hospital at the time of testing: less than 5% of the incident cases of psychosis identified by NIFEPS were treated in the community for their first episode. To ensure that concurrent clinical data would be available, testing occurred at the first date agreed by the participant after clinical interview. Symptoms were assessed using the Positive and Negative Syndrome Scale (PANSS) ^{Reference Kay, Opler and Lindenmayer9} and the Beck Depression Inventory (BDI). ^{Reference Beck, Steer and Brown10}

The neurocognitive battery measured a number of neuropsychological domains: global intellectual functioning (Wechsler Abbreviated Scale of Intelligence (WASI) two-test version: vocabulary and matrix reasoning); ^{Reference Wechsler11} premorbid IQ (National Adult Reading Test (NART)); ^{Reference Nelson12} verbal attention and working memory (Digit Span Forward/Backward from the Wechsler Memory Scale); ^{Reference Wechsler13} spatial attention (Corsi Block-tapping Test); ^{Reference Milner14} verbal memory (paired associates from the Wechsler Memory Scale); ^{Reference Wechsler13} visual memory (Rey Osterrieth Complex Figure Test with a recall trial at 30 min); ^{Reference Meyers and Meyers15} response inhibition and attentional set-shifting (Hayling and Brixton Tests); ^{Reference Burgess and Shallice16} verbal fluency (Controlled Oral Word Association (COWA) test); ^{Reference Lezak, Howieson, Loring, Hannay and Fischer17} language (Perin's Spoonerisms); ^{Reference Perin18} and callosal function (The Tactile Finger Localisation Test). ^{Reference Satomi, Kinoshita and Hirakawa19}

Statistical analysis

Data analysis was conducted using SPSS V.11.5 (SPSS Inc., 2002; PC version). Neuropsychological variables were assessed for normality (skewing and kurtosis). Extreme values were identified using boxplots and truncated to three standard deviations from the mean. To reduce positive skewing, a square root transformation was applied to Hayling error scores and participants' ages were log (ln) transformed. Missing data points for neuropsychological data were then imputed ^{Reference Schafer20} using NORM V.2.03, PC version (m = 3). The resulting test scores were standardised to the performance of the control group (z scored) and error scores were sign changed to provide a standard metric for comparison across tests.

Analysis of covariance (ANCOVA) was used to examine differences in test performance, with group (schizophrenia, bipolar disorder and controls) as a between-participants factor and test scores entered as dependent variables, with age entered as a covariate where it contributed to the model (groups differed significantly on this demographic variable). Post hoc Bonferroni-corrected pair-wise comparisons were used to detail group differences. Significant differences on clinical and general cognitive performance variables were then identified to determine potential explanatory covariates for consideration in further analyses. Descriptive statistics for these variables are shown in Table 1.

Table 1

Characteristics of bipolar disorder and schizophrenia groups relative to healthy comparison participants

	Controls	Bipolar disorder	Schizophrenia	‘Preserved’ IQ schizophreniaa	‘Low’ IQ schizophreniaa	Comparisons
Total, n	67	32	46	23	23
Gender, male: n (%)	39 (58.20)	16 (50.00)	32 (69.60)	15 (65.2)	17 (73.9)
Age, years: mean (s.d.)	33.19 (10.95)	36.71 (9.31)	28.99 (9.92)	26.02 (6.19)	31.97 (12.01)	SCZ<BPD,** preserved IQ<low IQ*
WASI IQ, mean (s.d.)	105.72 (11.28)	98.81 (16.81)	88.87 (13.00)	98.35 (11.68)	79.39 (4.66)	controls>SCZ,** controls>BD,* BD>SCZ,** preserved IQ>low IQ**
Premorbid IQ, mean (s.d.)	107.74 (9.98)	103.29 (11.88)	98.58 (11.61)	102.25 (12.21)	94.95 (9.54)	controls>SCZ,** controls>BD,* preserved IQ>low IQ*
PANSS total positive, mean (s.d.)		17.39 (5.51)	17.70 (6.06)	18.09 (5.21)	17.32 (6.88)
PANSS excitement—mania, mean (s.d.)b		6.09 (1.67)	5.96 (2.34)	5.78 (2.11)	6.13 (2.59)
PANSS total negative, mean (s.d.)		9.81 (3.25)	15.65 (5.1)	14.81 (4.58)	16.45 (5.54)	SCZ>BD**
BDI, mean (s.d.)		9.97 (10.83)	17.04 (10.29)	16.14 (10.07)	17.23 (11.07)	SCZ>BD**

The influence of potential explanatory covariates for group differences (i.e. those variables on which groups differed) were examined sequentially in a series of ANCOVAs to determine the extent to which each of these factors contributed to the neurocognitive differences observed between the groups. Given justifiable concerns about covarying for IQ where IQ impairment is considered intrinsic to a disorder, we then examined the effect of a minimal v. a sizeable global cognitive impairment on the overall shape of the schizophrenia group's neurocognitive profile relative to the bipolar disorder group, who showed preserved IQ. A ‘low IQ’ schizophrenia group and a ‘preserved IQ’ schizophrenia group were created for this purpose using a median split of the current IQ variable in people with schizophrenia (median 87, producing a split of the schizophrenia group at approximately 1 standard deviation below the expected average IQ for the general population). Neurocognitive performance according to these ‘preserved’ and ‘low’ IQ designations were then plotted and differences between groups were again compared using ANCOVA (age entered as a covariate where it contributed to the model).

Finally, as the covariates WASI IQ, negative symptoms and depression were correlated in this sample (IQ and negative symptoms, r = −0.3, P<0.01; IQ and depression, r = −0.18, P = 0.06; depression and negative symptoms, r = 0.32, P<0.01), multiple regression was used to determine the unique variance in neurocognitive performance attributable to each of the factors considered in these analyses.

Selective and general cognitive impairments

The non-specific profile of multiple cognitive deficits observed in schizophrenia has been the subject of considerable debate in the literature. ^{Reference Mohamed, Paulsen, O'Leary, Arndt and Andreasen22,Reference Bilder, Goldman, Robinson, Reiter, Bell and Bates23} It has been suggested that a non-specific deficit pattern is indicative of either diffuse dysfunction in cortical circuitry or disturbances in key systems that modulate broadly distributed neural networks. ^{Reference Mohamed, Paulsen, O'Leary, Arndt and Andreasen22} Comparisons between the bipolar disorder group, the schizophrenia ‘preserved’ and ‘low’ IQ groups contribute to this debate. The bipolar disorder group had impairments in many selective domains of neurocognitive function against a background of preserved general cognitive functioning. The shape of their cognitive profile was also very similar to that of the ‘preserved IQ’ schizophrenia group. The most pronounced deficits in both bipolar disorder and schizophrenia were in visual and verbal memory, executive set shifting and language functioning. These clinically significant selective deficits in neurocognitive function do not appear to be artefacts of a global cognitive impairment or negative symptoms. Rather these particular deficits possibly underlie psychosis per se, reflecting abnormalities in a common pathway at a systems level that is evident early in the course of the psychoses. Their utility as putative endophenotypes may not be specific to schizophrenia and it is reasonable to argue that cohorts of people with psychosis could be examined as a single set of interest to improve statistical power in genetic studies that make use of these endophenotypes.

However, as the majority of participants were unwell at the time of testing, it is essential to examine these neurocognitive domains longitudinally in the NIFEPS cohort to determine the extent to which these are ‘trait’ or ‘state’ deficits. It might be hypothesised that the group with bipolar disorder examined in this study showed a fair degree of normalisation in cognition as they became progressively euthymic. Of particular interest would be comparisons between individuals with bipolar disorder and those individuals with ‘preserved IQ’ and schizophrenia following a more lengthy period of treatment and improvements in symptoms. It is possible that what makes schizophrenia unique from a neurocognitive perspective is not the precise pattern or magnitude of selective deficits observed cross-sectionally, but the failure to improve on specific selective domains with the resolution of psychotic symptoms regardless of whether IQ is ‘preserved’ at first episode or not.

Implications for the use of diagnostic criteria

A related issue is how these data affect the use of diagnostic classification in neurocognitive research of the psychoses. It could be tentatively argued with reference to the limitation outlined above that people with schizophrenia and a ‘preserved’ IQ are more suitably grouped with people with bipolar disorder, or alternatively, that individuals with psychosis should be divided into separate groups according to whether they fall above or below 1 standard deviation of the expected mean IQ for the general population for example. Comparing ‘preserved IQ’ diagnostic groups, or clustering these ‘preserved IQ’ participants according to specific symptoms for comparative purposes, may shed greater light on how the severity of positive, affective and, potentially, negative symptoms contribute to selective neurocognitive impairments and functional outcome in people with psychosis in the absence of the highly pervasive influence of low IQ within a cohort. Given recent meta-analyses ^{Reference Robinson, Thompson, Gallagher, Goswami, Young and Ferrier28,Reference Torres, Boudreau and Yatham29} that identify impairments in people with euthymic bipolar disorder in the domains of attention, executive functioning and memory, it is possible that the existing literature on neurocognitive functioning in individuals with bipolar disorder may prove of some relevance to individuals with schizophrenia and a relatively preserved IQ. It might also be speculated that grouping people with psychosis according to ‘preserved’ and ‘low’ IQ may be of particular use when prospectively conducting a comparative examination of the longitudinal course of the psychoses, and when retrospectively examining aspects of the neurodevelopmental hypothesis of schizophrenia. However, the issues of ‘potential’ v. ‘achieved’ IQ of people with schizophrenia in adolescence, and the decline in IQ that can occur following the onset of illness, would need to be carefully considered before adopting these approaches.

Confounding effects of medication

A second limitation to the conclusions from these analyses is the potential confounding effects of medication. Given the fact that almost all of the participants were medicated at the time of testing, most receiving antipsychotics, it was not possible to fully gauge the contribution this factor made to their neurocognitive functioning. Furthermore, medications may have influenced the neurocognitive performance of the schizophrenia and bipolar disorder groups differently. Undoubtedly, such issues can only be addressed through controlled clinical trials, and studies examining cognition in individuals who have never received medication. However, the latter have consistently demonstrated that people who are antipsychotic naive show similar impairments to those on medication. Brickman et al ^{Reference Brickman, Buchsbaum, Bloom, Bokhoven, Paul-Odouard and Haznedar24} compared 29 never-medicated adolescents with a first-episode psychosis with 17 matched controls and reported that those with psychosis significantly underperformed on all measures, but were most impaired on tests of memory, attention, and executive functioning. Saykin and colleagues ^{Reference Saykin, Shtasel, Gur, Kester, Mozley and Stafiniak25} also reported that antipsychotic-naive patients with a schizophrenia-spectrum first-episode psychosis (n = 37) had a near identical profile of neurocognitive deficits to previously treated but currently unmedicated individuals with schizophrenia (n = 67), both groups showing generalised impairment and selective deficits in attention, memory and executive functioning. Similar conclusions were also reached by Hill et al ^{Reference Hill, Keshavan, Thase and Sweeney3} in a large study that compared participants with a schizophrenia-spectrum first-episode psychosis (n = 86) with participants with a first episode of psychotic depression (n = 14) and healthy volunteers (n = 81). Medication is therefore highly unlikely to account for the neuropsychological deficits observed in our study of people with first-episode psychosis.

Stability of diagnoses

A third limitation relates to the stability of individuals' diagnoses. It is recognised that diagnosing people during their first episode is difficult and it is possible that individuals classified as having a bipolar disorder may change to schizophrenia (and, although less likely, vice versa). Paradoxically, this, as well as problems related to the availability of different diagnostic classification systems, forms part of the rationale for abandoning diagnosis as a variable in research studies. In this study, stability of diagnoses was good at 1 year as determined by operational criteria (OPCRIT) ^{Reference McGuffin, Farmer and Harvey30} ICD–10: diagnostic stability was 89% for the schizophrenia group (0% switch to bipolar disorder, 5% switch to psychotic depression) and 72% for the bipolar disorder group (16% switch to schizophrenia), which is similar or better than that reported in the literature. ^{Reference Chen, Swann and Johnson31,Reference Mihalopoulos, McGorry, Roberts and McFarlane32}

Implications for research and practice

Overall, this study demonstrates that neurocognitive impairments in memory, executive functioning and language are common to people with first-episode psychosis and with further study, these neurocognitive impairments may prove useful as endophenotypic markers for psychosis in general. Furthermore, consideration of the degree of global cognitive impairment within cohorts is critical when attempting to understand both diagnostic differences between psychosis groups and patterns of selective neurocognitive impairment in people with schizophrenia.