General intellectual ability does not explain the general deficit in schizophrenia

Abstract

Patients with schizophrenia demonstrate a generalized deficit across multiple cognitive domains. However, it is unknown whether this deficit is largely due to lower intelligence, or if there is an impact of schizophrenia which cannot be accounted for by measures of general intellectual ability (GIA). We created four IQ-matched strata of equal width between 89 healthy volunteers (HC) and 77 patients with schizophrenia (SZ) who had very similar IQ and reading scores within each stratum, then compared each stratum's performance on the MATRICS Consensus Cognitive Battery (MCCB). We hypothesized that any patient impairment on the MCCB after matching on IQ would be evidence that GIA does not fully explain the general deficit seen in schizophrenia. We found that patients showed evidence of greater neuropsychological impairment than what would be expected based solely on their IQ and reading ability scores. Further, this deficit was stronger in some cognitive domains than others, namely, processing speed and social cognition. These results suggest the presence of a distinction between GIA and generalized neuropsychological impairment that was consistent in magnitude across all patients, regardless of IQ.

Introduction

Meta-analyses provide overwhelming evidence that patients with schizophrenia demonstrate marked deficits across cognitive domains (Heinrichs and Zakzanis, 1998, Fioravanti et al., 2005, Dickinson et al., 2007, Reichenberg and Harvey, 2007, Reichenberg, 2010). While there is variability in the extent of impairment, evidence suggests that impairment is generalized across the cognitive operations assessed by widely used clinical neuropsychological measures. Further, it appears that the extent of impairment across these domains (i.e., attention, processing speed, working memory, etc.), is highly intercorrelated. Dickinson et al. (2008) used structural equation modeling to demonstrate that 64% of the between-group variance in neuropsychological performance between healthy controls and individuals with schizophrenia is shared on a common general deficit factor, with more specific deficits accounting for very little additional between-group variance (Dickinson et al., 2004, Dickinson et al., 2008).

With evidence of a generalized deficit across cognitive domains, the question arises whether the “general deficit” might simply be a reflection of a reduction in general intellectual ability (GIA), i.e., intelligence. Indeed, IQ measures are typically highly correlated with neuropsychological performance. For example, in a sample of 117 individuals with schizophrenia (SZ), WASI-estimated IQ scores correlated with the composite score from the MATRICS Consensus Cognitive Battery (MCCB, Kern et al., 2008, Nuechterlein et al., 2008), r = .733, p < .001 with a very similar correlation observed in a sample of 77 healthy controls (HCs), r = .695, p < .001 (August et al., 2012). These substantial correlations are noteworthy because the MCCB was deliberately composed of measures particularly impaired in schizophrenia and/or particularly important for functional outcome. Thus, one would expect to see a schizophrenia deficit “signal” in MCCB performance that extends beyond GIA. We speculate that across the WASI and MCCB there are two “pools” of variance: 1) a pool of variance associated with GIA reflected in the high correlation of the two measures, and 2) a pool of variance associated with the impact of schizophrenia on more discrete aspects of cognitive function that are captured on the MCCB which cannot be accounted for by GIA.

We took two approaches to this issue. First, we compared the MCCB performance of healthy volunteers and patients with schizophrenia who had very similar WASI IQ scores. If IQ accounts for neuropsychological performance across groups, the IQ-matched groups should show similar levels of performance on the MCCB. Alternatively, any patient impairment on the MCCB, after matching on IQ, would be evidence that the “general deficit” and GIA are not synonymous. In addition, we performed the same matched group approach using measures of single word reading which are thought to index “premorbid” ability (Spreen and Strauss, 1998, Lezak et al., 2004). With both the WASI and reading measures, this approach addresses the question of whether patients are more impaired than they “should” be for their level of reading and IQ performance, and allows for a quantitative estimate of how far patients deviate from the level that would be expected had they not become ill. We examined these questions by creating groups that ranged from low to high levels of GIA to provide additional information about whether patients who have higher levels of cognitive ability are spared the neuropsychological impairments that have most frequently been documented in samples with average–low average levels of GIA. Second, we used an ANCOVA approach that provides further information on whether patients show greater impairments in some domains than others after controlling for the role of GIA.