Introduction
Evidence has accumulated to indicate that schizophrenia is, in part, a neurodevelopmental disorder (Murray & Lewis, Reference Murray and Lewis1987; Weinberger, Reference Weinberger1987) characterized by abnormal functioning during childhood and adolescence (Niemi et al. Reference Niemi, Suvisaari, Tuulio-Henriksson and Lönnqvist2003; Schenkel & Silverstein, Reference Schenkel and Silverstein2004). Converging evidence from prospective longitudinal studies of population cohorts, prospective studies of individuals at elevated risk of schizophrenia because they have a family history of the illness, and ‘follow-back’ studies of adults with schizophrenia suggest that cognitive and motor dysfunctions precede the onset of schizophrenia. However, past literature reviews of cognitive functioning among children who develop schizophrenia/schizophrenia spectrum disorders (SSD) in adulthood have been limited in several respects: (i) they have focused exclusively on individuals with affected relatives (Niemi et al. Reference Niemi, Suvisaari, Tuulio-Henriksson and Lönnqvist2003; Keshavan et al. Reference Keshavan, Kulkarni, Bhojraj, Francis, Diwadkar, Montrose, Seidman and Sweeney2010) or on population and/or birth cohorts (MacCabe, Reference MacCabe2008; Welham et al. Reference Welham, Isohanni, Jones and McGrath2009a); (ii) they have not followed samples into adulthood and assessed them for schizophrenia/SSD; or (iii) they have included studies with estimated measures of pre-morbid intellectual functioning assessed in adulthood when the participant already had a diagnosis of schizophrenia (Schenkel & Silverstein, Reference Schenkel and Silverstein2004). Thus, it remains unclear whether cognitive and motor deficits constitute robust antecedents of schizophrenia.
During childhood, individuals who subsequently developed schizophrenia or SSD, compared to those who did not, were characterized by lower IQ (Woodberry et al. Reference Woodberry, Giuliano and Seidman2008) and poorer motor function (Walker et al. Reference Walker, Savoie and Davis1994; Rosso et al. Reference Rosso, Bearden, Hollister, Gasperoni, Sanchez, Hadley and Cannon2000; Cannon et al. Reference Cannon, Caspi, Moffitt, Harrington, Taylor, Murray and Poulton2002, Reference Cannon, Moffitt, Caspi, Murray, Harrington and Poulton2006; Schiffman et al. Reference Schiffman, Walker, Ekstrom, Schulsinger, Sørensen and Mednick2004). Additional cognitive decline prior to the onset of the prodromal phase of schizophrenia has been reported (Fuller et al. Reference Fuller, Nopoulos, Arndt, O'Leary, Ho and Andreasen2002; Osler et al. Reference Osler, Lawlor and Nordentoft2007). Academic achievement has also distinguished children who later developed schizophrenia/SSD, although findings have differed according to subject and assessment type (Watt & Lubensky, Reference Watt and Lubensky1976; Jones et al. Reference Jones, Rodgers, Murray and Marmot1994; Crow et al. Reference Crow, Done and Sacker1995; Fuller et al. Reference Fuller, Nopoulos, Arndt, O'Leary, Ho and Andreasen2002; Bilder et al. Reference Bilder, Reiter, Bates, Lencz, Szeszko, Goldman, Robinson, Lieberman and Kane2006; MacCabe et al. Reference MacCabe, Lambe, Cnattingius, Torrång, Björk, Sham, David, Murray and Hultman2007). Furthermore, five studies reported no differences in pre-morbid academic performance (Isohanni et al. Reference Isohanni, Auml Rvelin, Nieminen, Jones, Rantakallio, Jokelainen and Isohanni1998; Cannon et al. Reference Cannon, Jones, Huttunen, Tanskanen, Huttunen, Rabe-Hesketh and Murray1999; Isohanni et al. Reference Isohanni, Jarvelin, Jones, Jokelainen and Isohanni1999; Helling et al. Reference Helling, Öhman and Hultman2003; Ang & Tan, Reference Ang and Tan2004), which may reflect differences in the education systems characterizing the study cohorts. Taken together, the extant literature suggests that individuals who develop schizophrenia present poorer cognitive abilities in childhood than those who never develop schizophrenia/SSD. To date, little is known about the age at which these deficits emerge or their specific nature.
Two previous, widely cited, meta-analyses evaluated IQ among individuals who subsequently developed schizophrenia, and both yielded medium-sized deficits (Aylward et al. Reference Aylward, Walker and Bettes1984; Woodberry et al. Reference Woodberry, Giuliano and Seidman2008). The present meta-analysis differs from the more recent of these, by Woodberry and colleagues, in several ways. First, several studies that were included in that meta-analysis reported the results of IQ assessments completed at multiple ages spanning a broad age range of 3–19 years (Albee et al. Reference Albee, Lane and Reuter1964; Watt & Lubensky, Reference Watt and Lubensky1976; Jones et al. Reference Jones, Rodgers, Murray and Marmot1994; Ott et al. Reference Ott, Spinelli, Rock, Roberts, Amminger and Erlenmeyer-Kimling1998; Cannon et al. Reference Cannon, Bearden, Hollister, Rosso, Sanchez and Hadley2000, Reference Cannon, Caspi, Moffitt, Harrington, Taylor, Murray and Poulton2002; Seidman et al. Reference Seidman, Buka, Goldstein and Tsuang2006). An overall unweighted mean effect size was calculated for each of these studies spanning multiple assessments and a broad age range, thereby providing a less robust estimate of pre-morbid IQ than might be achieved by using a single assessment completed during childhood/adolescence. Second, the previous meta-analysis included studies of symptomatic, help-seeking individuals meeting inclusion criteria for treatment in an intervention programme for persons at ultra-high risk for psychosis (Brewer et al. Reference Brewer, Francey, Wood, Jackson, Pantelis, Phillips, Yung, Anderson and McGorry2005; Lencz et al. Reference Lencz, Smith, McLaughlin, Auther, Nakayama, Hovey and Cornblatt2006), and studies examining IQ among young adults (Lubin et al. Reference Lubin, Gieseking and Williams1962; Zammit et al. Reference Zammit, Allebeck, David, Dalman, Hemmingsson, Lundberg and Lewis2004; Reichenberg et al. Reference Reichenberg, Weiser, Rapp, Rabinowitz, Caspi, Schmeidler, Knobler, Lubin, Nahon, Harvey and Davidson2005; Kremen et al. Reference Kremen, Lyons, Boake, Xian, Jacobson, Waterman, Eisen, Goldberg, Faraone and Tsuang2006; Whyte et al. Reference Whyte, Brett, Harrison, Byrne, Miller, Lawrie and Johnstone2006). Given that intellectual deficits have been reported to increase in magnitude with the onset of psychosis (Rabinowitz et al. Reference Rabinowitz, Reichenberg, Weiser, Mordechai, Kaplan and Davidson2000; Gunnell et al. Reference Gunnell, Harrison, Rasmussen, Fouskakis and Tynelius2002; Caspi et al. Reference Caspi, Reichenberg, Weiser, Rabinowitz, Kaplan, Knobler, Davidson-Sagi and Davidson2003), the effect sizes of the difference between participants who did and who did not develop schizophrenia in these studies may reflect some early prodromal disease process rather than an antecedent of schizophrenia. Third, although the previous meta-analysis included subanalyses that examined IQ by narrower age bands, only three studies were included in the meta-analysis that assessed participants aged 13 years or younger. Finally, unlike both previous meta-analyses that focused solely on IQ (Aylward et al. Reference Aylward, Walker and Bettes1984; Woodberry et al. Reference Woodberry, Giuliano and Seidman2008), the present meta-analysis includes examination of additional domains of academic achievement and motor functioning.
The aim of the present meta-analyses was to examine IQ, motor function and academic achievement in children and young adolescents (aged ⩽16 years) who subsequently developed schizophrenia/SSD. A recent review of studies on individuals at ultra-high risk for psychosis indicated a typical age of onset of basic prodromal symptoms of greater than 16 years (Ruhrmann et al. Reference Ruhrmann, Schultze-Lutter and Klosterkötter2010). By restricting the meta-analyses to studies of children aged 16 years or younger, the present meta-analyses aimed to determine whether deficits in IQ, motor function and academic achievement are present before the typical age of onset of the prodrome.
Method
Sample
Meta-analyses were conducted to identify the effect sizes of differences in scores obtained on cognitive and motor performance measures by individuals aged 16 years or younger who subsequently developed schizophrenia compared to those who did not. Articles were identified through literature searches conducted in PubMed/Medline and PsycINFO, using the keywords ‘schizophrenia’ and ‘IQ’ or ‘intelligence’ or ‘motor’ or ‘school’ or ‘scholastic’ and ‘premorbid’ or ‘prospective’ or ‘cohort’ or ‘high risk’. References from articles and relevant literature reviews were also examined for possible inclusion in meta-analyses. Inclusion criteria were: (1) written in English; (2) published or unpublished prospective investigations of birth cohorts or genetic high-risk samples, or follow-back investigations of population samples; (3) objective measures of cognitive or motor function when participants were aged 16 or younger; (4) results provided for cohort members who did and who did not develop schizophrenia or an SSD later in life; and (5) sufficient data to calculate effect sizes.
The initial literature search by the first author identified 2623 studies, of which 34 fulfilled the inclusion criteria. A co-author independently reviewed these studies to verify that inclusion criteria were met. Among the 34 studies, 10 contained samples that overlapped (Lane & Albee, Reference Lane and Albee1963, Reference Lane and Albee1968; Albee et al. Reference Albee, Lane and Reuter1964; Cannon et al. Reference Cannon, Bearden, Hollister, Rosso, Sanchez and Hadley2000, Reference Cannon, Caspi, Moffitt, Harrington, Taylor, Murray and Poulton2002, Reference Cannon, Moffitt, Caspi, Murray, Harrington and Poulton2006; Niendam et al. Reference Niendam, Bearden, Rosso, Sanchez, Hadley, Nuechterlein and Cannon2003; Seidman et al. Reference Seidman, Buka, Goldstein and Tsuang2006). To avoid multiple entries on the same sample, only data from the study containing the largest number of participants were analysed (Albee et al. Reference Albee, Lane and Reuter1964; Cannon et al. Reference Cannon, Bearden, Hollister, Rosso, Sanchez and Hadley2000, Reference Cannon, Caspi, Moffitt, Harrington, Taylor, Murray and Poulton2002; Seidman et al. Reference Seidman, Buka, Goldstein and Tsuang2006). Ten studies reported assessments of participants at multiple ages ranging from 3 to 19 years (Watt & Lubensky, Reference Watt and Lubensky1976; Jones et al. Reference Jones, Rodgers, Murray and Marmot1994; Crow et al. Reference Crow, Done and Sacker1995; Ott et al. Reference Ott, Spinelli, Rock, Roberts, Amminger and Erlenmeyer-Kimling1998; Cannon et al. Reference Cannon, Bearden, Hollister, Rosso, Sanchez and Hadley2000, Reference Cannon, Caspi, Moffitt, Harrington, Taylor, Murray and Poulton2002; Rosso et al. Reference Rosso, Bearden, Hollister, Gasperoni, Sanchez, Hadley and Cannon2000; Ang & Tan, Reference Ang and Tan2004; Bilder et al. Reference Bilder, Reiter, Bates, Lencz, Szeszko, Goldman, Robinson, Lieberman and Kane2006; Welham et al. Reference Welham, Scott, Williams, Najman, Bor, O'Callaghan and McGrath2009b). From these studies, only the results from a single assessment completed when participants were between 4 and 14 years old were included in the meta-analyses. Seven studies reported insufficient data to calculate an effect size (Ambelas, Reference Ambelas1992; Crow et al. Reference Crow, Done and Sacker1995; Cannon et al. Reference Cannon, Jones, Huttunen, Tanskanen, Huttunen, Rabe-Hesketh and Murray1999, Reference Cannon, Moffitt, Caspi, Murray, Harrington and Poulton2006; Erlenmeyer-Kimling et al. Reference Erlenmeyer-Kimling, Rock, Roberts, Janal, Kestenbaum, Cornblatt, Adamo and Gottesman2000; Fuller et al. Reference Fuller, Nopoulos, Arndt, O'Leary, Ho and Andreasen2002; Walker et al. Reference Walker, McConville, Hunter, Deary and Whalley2002). Additional data from authors were obtained for all but two studies (Ambelas, Reference Ambelas1992; Walker et al. Reference Walker, McConville, Hunter, Deary and Whalley2002), but data from two studies remained insufficiently detailed to satisfy the inclusion criteria (Erlenmeyer-Kimling et al. Reference Erlenmeyer-Kimling, Rock, Roberts, Janal, Kestenbaum, Cornblatt, Adamo and Gottesman2000; Fuller et al. Reference Fuller, Nopoulos, Arndt, O'Leary, Ho and Andreasen2002). Two other papers reported similar data (Isohanni et al. Reference Isohanni, Auml Rvelin, Nieminen, Jones, Rantakallio, Jokelainen and Isohanni1998, Reference Isohanni, Jarvelin, Jones, Jokelainen and Isohanni1999); only data in the 1998 publication were included in the meta-analyses. Two studies that reported IQ scores for participants aged between 8 and 20 years were excluded (Bower et al. Reference Bower, Shellhamer and Daily1960; Sørensen et al. Reference Sørensen, Mortensen, Parnas and Mednick2006). Another study, which presented results for performance IQ between members of a rubella-exposed birth cohort who subsequently developed schizophrenia and those who did not, was also excluded from the meta-analysis (Brown et al. Reference Brown, Cohen, Harkavy-Friedman, Babulas, Malaspina, Gorman and Susser2001).
Meta-analyses were performed using the results from the 23 studies that fulfilled the inclusion criteria. The results were categorized into four domains of cognitive and motor function. Table 1 presents details of each study included in the meta-analyses, with a description of the sample, participant age at assessment, the test instrument used, and effect sizes denoting the difference in performance between the participants who subsequently developed schizophrenia/SSD compared to those who did not. Across all 23 studies, the age of the participants at the time of assessment ranged from 2 to 16 years. Twenty-one of the 23 studies included males and females whereas two examined only males (Ang & Tan, Reference Ang and Tan2004; Osler et al. Reference Osler, Lawlor and Nordentoft2007). The comparison groups varied widely across studies, and were described as classmates, child psychiatric patients with no adult mental disorder, members of birth cohorts who did not develop schizophrenia/SSD, members of birth cohorts who did not develop any major mental disorder, and participants with or without a family history of schizophrenia/SSD.
Table 1. Study details and effect sizes for meta-analyses
SSD, Schizophrenia spectrum disorder; WISC-R, Wechsler Intelligence Scale for Children – Revised.
a Effect sizes were estimated using Cohen's d, obtained using sample sizes, means and standard deviations for a group who later developed schizophrenia or an SSD and a comparison group, except for the following: (i) IQ: for Offord (Reference Offord1974), the results were presented across gender, so data were collapsed and overall means and standard deviations were used; for Watt & Lubensky (Reference Watt and Lubensky1976), the effect size was computed from the sample size and t statistic; for Jones et al. (Reference Jones, Rodgers, Murray and Marmot1994), effect sizes were taken from the Woodberry et al. (Reference Woodberry, Giuliano and Seidman2008) meta-analysis; for Crow et al. (Reference Crow, Done and Sacker1995), effect size was estimated from the sample size and f statistic; for Welham et al. (Reference Welham, Isohanni, Jones and McGrath2009b), effect size was calculated by converting β to t statistics (b/seB) with effect size derived from t statistic and sample size for both males and females. A mean weighted effect size was then calculated based on sample size by gender. (ii) Motor Function: for Rosso et al. (Reference Rosso, Bearden, Hollister, Gasperoni, Sanchez, Hadley and Cannon2000), the odds ratio was transformed into a Cohen's d using a method outlined by (Chinn, Reference Chinn2000). (iii) Academic Performance: Mathematics: for both Jones et al. (Reference Jones, Rodgers, Murray and Marmot1994) and Helling et al. (Reference Helling, Öhman and Hultman2003), sample size and f statistics were used; and for Crow et al. (Reference Crow, Done and Sacker1995), t statistic was calculated from degrees of freedom and p value given in paper; effect size was then estimated from sample size and t statistic.
b Positive values indicate better performance in the comparison group.
c Sample size for males and females were taken from measures of attentional dysfunction because not available for Raven's Standard Progressive Matrices Test.
Statistical analyses
Meta-analyses were conducted with Stata version 10 (Stata Corporation, USA) using a random effects model (Dersimonian & Laird, Reference Dersimonian and Laird1986) that assumes that the effects being investigated in a set of studies are a random sample drawn from a population of possible effect sizes. Meta-analyses were performed on difference scores for each domain of functioning, comparing participants who developed schizophrenia/SSD to those who did not. Difference scores were standardized by calculating Cohen's d effect sizes (Cohen, Reference Cohen1988) and interpreted according to effect size indices of ‘small (0.2)’, ‘medium (0.5)’ and ‘large (0.8)’ (Cohen, Reference Cohen1992). The summary effect sizes for each domain were the standardized mean differences (SMDs), weighted by the precision of the SMD. For each SMD, a z value and a significance level provided an indication of the two-sided statistical significance of the association at the 95% probability level. For the IQ domain, the effect size from one study was an extreme outlier (Woodberry et al. Reference Woodberry, Giuliano and Seidman2008), so the analysis was conducted with and without the inclusion of this study (Amminger et al. Reference Amminger, Schlögelhofer, Lehner, Ott, Friedrich and Aschauer2000).
The significance and the magnitude of heterogeneity across studies were calculated using the Q statistic and I 2 statistic. Where there was significant heterogeneity within a domain, and where there were sufficient studies to provide adequate statistical power (i.e. for IQ only), effect size moderators were examined. Three potential moderator variables were examined: comparison group, IQ assessment instrument, and disorder outcome. Comparison group (matched comparison group or unselected cohort) was included as it had been reported to be a significant source of heterogeneity in a previous meta-analysis on IQ (Woodberry et al. Reference Woodberry, Giuliano and Seidman2008). An instrument used to assess IQ (i.e. Wechsler Intelligence Scales or other test) was included as different types of IQ tests, particularly tests that are older, may provide variable estimates of IQ (Sattler, Reference Sattler2001). We also examined disease outcome (i.e. schizophrenia or SSD) based on the rationale that individuals who develop schizophrenia may differ from those who develop SSD. For each variable, a regression model was estimated using an unrestricted maximum likelihood model. Publication bias was assessed graphically and statistically using published methods (Begg & Mazumdar, Reference Begg and Mazumdar1994; Egger et al. Reference Egger, Davey Smith, Schneider and Minder1997). In domains with five studies or less, publication bias could not be explored (Sutton et al. Reference Sutton, Duval, Tweedie, Abrams and Jones2000).
Results
IQ
A meta-analysis of the 13 studies included in the IQ domain indicated that youth aged 16 years or younger who subsequently developed schizophrenia/SSD obtained lower IQ scores than youth who did not develop these disorders. As illustrated in Fig. 1, a medium effect size was obtained [SMD 0.51, 95% confidence interval (CI) 0.38–0.65, z=7.51, p<0.001]. Significant heterogeneity was detected across studies (Q=26.55, df=12, p<0.05/I 2=54.8%), but neither comparison group, IQ measure nor diagnostic outcome explained the heterogeneity. No publication bias was detected. After removing one study from the meta-analysis due to an effect size that was an outlier to the group (Amminger et al. Reference Amminger, Schlögelhofer, Lehner, Ott, Friedrich and Aschauer2000), the observed effect size remained of medium magnitude (SMD 0.49, 95% CI 0.37–0.61, z=8.03, p<0.001). Significant heterogeneity was detected (Q=20.25, df=11, p<0.05/I 2=45.7%). Again, neither comparison group, IQ measure nor diagnostic outcome was associated with heterogeneity.
Fig. 1. Forest plot for IQ.
The meta-analysis was repeated after excluding two studies that assessed participants between the ages of 14 and 16 years (Amminger et al. Reference Amminger, Schlögelhofer, Lehner, Ott, Friedrich and Aschauer2000; Welham et al. Reference Welham, Scott, Williams, Najman, Bor, O'Callaghan and McGrath2009b). All participants in the remaining 11 studies were aged 13 years or younger. This analysis yielded an effect size that was similar in magnitude to that calculated for participants aged 16 years or younger (SMD 0.51, 95% CI 0.38–0.64, z=7.69, p<0.001). As before, significant heterogeneity was detected (Q=19.11, df=10, p<0.05/I 2=47.7%). Again, heterogeneity was not associated with the type of comparison group, the measure of IQ or the outcome diagnosis of schizophrenia or SSD. No publication bias was detected.
Motor function
Of the four studies included in the motor function domain, the results from the meta-analysis showed that individuals aged 16 years or younger who subsequently developed schizophrenia/SSD, as compared to those who did not, displayed significant deficits in motor function (see Fig. 2) that were moderate in size (SMD 0.56, 95% CI 0.38–0.74, z=6.25 p<0.001). No significant heterogeneity was detected across studies (Q=1.87, df=3, p=0.60/I 2=0.0%).
Fig. 2. Forest plot for motor function.
Academic Achievement
General
Five studies examined General Academic Achievement among youth aged 16 years or younger who subsequently developed schizophrenia/SSD, and the results indicated poorer overall academic achievement compared to youth who did not later develop schizophrenia/SSD (see Fig. 3). However, the effect size of the group difference was small and non-significant (SMD 0.25, 95% CI −0.03 to 0.53, z=1.74, p=0.08). Significant heterogeneity in the results was detected (Q=40.72, df=4, p<0.001/I 2=90.2%), but could not be examined further given the limited number of studies comprising this domain.
Fig. 3. Forest plot for Academic Achievement: General.
Mathematics
The results of the meta-analysis of four studies showed that youth aged 16 years and younger who later developed schizophrenia/SSD, as compared to those who did not, achieved more poorly on tests of mathematics. However, as indicated in Fig. 4, the effect size of the difference between the two groups was small and non-significant (SMD 0.21, 95% CI −0.09 to 0.51, z=1.40, p=0.16). Significant heterogeneity was detected (Q=7.96, df=3, p<0.05/I 2=62.3%), but could not be examined further because of an insufficient number of studies within the domain.
Fig. 4. Forest plot for Academic Achievement: Mathematics.
Discussion
To our knowledge, these are the first meta-analyses examining both cognitive and motor performance among youth aged 16 years or younger who later developed schizophrenia/SSD. The meta-analyses demonstrate that participants who subsequently developed schizophrenia/SSD displayed lower IQ and poorer motor function by age 16 than individuals who did not develop these disorders. Furthermore, there were sufficient studies to conduct a meta-analysis that showed that the deficit in IQ was present by age 13. By contrast, overall academic achievement and performance on tests of mathematics did not significantly distinguish those who subsequently developed schizophrenia/SSD from those who did not. These results extend previous findings by establishing that low IQ and impaired motor performance precede the prodrome and onset of illness.
Although significant heterogeneity was detected in the meta-analyses of IQ and the two domains of academic performance, the factors affecting heterogeneity could be examined only for IQ. Analyses indicated that heterogeneity reported for the IQ results was not due to the type of comparison group used (matched comparison or unselected cohort), the test used to assess IQ (Wechsler or other) or diagnostic outcome (schizophrenia or SSD).
The present meta-analyses obtained the same medium effect size for the difference in IQ by age 16 and by age 13 (d=0.51), between individuals who did and who did not develop schizophrenia/SSD later in life. The present IQ meta-analysis included results from five studies (Crow et al. Reference Crow, Done and Sacker1995; Amminger et al. Reference Amminger, Schlögelhofer, Lehner, Ott, Friedrich and Aschauer2000; Osler et al. Reference Osler, Lawlor and Nordentoft2007; Welham et al. Reference Welham, Scott, Williams, Najman, Bor, O'Callaghan and McGrath2009b; Sørensen et al. Reference Sørensen, Mortensen, Schiffman, Ekstrøm, Denenney and Mednick2010) that had not been included in previous meta-analyses (Aylward et al. Reference Aylward, Walker and Bettes1984; Woodberry et al. Reference Woodberry, Giuliano and Seidman2008). Nevertheless, the effect sizes obtained in the present analyses by age 16 and by age 13 are similar in magnitude to that reported in a previous meta-analysis (d=0.54) that included participants who were symptomatic or deemed to be at risk for psychosis (Woodberry et al. Reference Woodberry, Giuliano and Seidman2008). Thus, our meta-analyses indicate that, among individuals who later develop schizophrenia/SSD, a deficit in IQ is measurable by early to mid-adolescence.
Consistent with the findings of a previous meta-analysis (Woodberry et al. Reference Woodberry, Giuliano and Seidman2008), our current analysis indicated that the specific measure of IQ used did not account for significant heterogeneity in the results. Unlike the present meta-analysis, however, Woodberry et al. (Reference Woodberry, Giuliano and Seidman2008) reported that the type of comparison group (i.e. unselected cohort versus matched comparison group) accounted for significant effect-size heterogeneity. Their finding was attributed primarily to the inclusion of one study of 18-year-old military conscripts that used an army classification battery to assess IQ (Lubin et al. Reference Lubin, Gieseking and Williams1962).
It was not possible to examine all potential moderator variables in the meta-analysis of IQ. This was because of the relatively small number of studies that met inclusion criteria and the limited data available from each study. Examining a larger number of potential moderators may increase the likelihood of drawing false positive conclusions (Thompson & Higgins, Reference Thompson and Higgins2002). Given the evidence that the prevalence of schizophrenia is higher in males than in females (Aleman et al. Reference Aleman, Kahn and Selten2003; McGrath et al. Reference McGrath, Saha, Welham, El Saadi, MacCauley and Chant2004), it is possible that gender differences may have contributed to the heterogeneity of the results in the IQ domain, as reported in one previous meta-analysis (Aylward et al. Reference Aylward, Walker and Bettes1984), but not in another (Woodberry et al. Reference Woodberry, Giuliano and Seidman2008). Unfortunately, the data available for inclusion in the present study were insufficient to permit statistical analysis of gender differences. Five studies included in the present meta-analysis reported the prevalence of schizophrenia separately by gender, and all but one (Welham et al. Reference Welham, Scott, Williams, Najman, Bor, O'Callaghan and McGrath2009b) found higher rates among males than females (Offord, Reference Offord1974; Jones et al. Reference Jones, Rodgers, Murray and Marmot1994; Cannon et al. Reference Cannon, Bearden, Hollister, Rosso, Sanchez and Hadley2000; Seidman et al. Reference Seidman, Buka, Goldstein and Tsuang2006). In these studies, the effect sizes of the differences in IQ of participants who did and who did not develop schizophrenia/SSD ranged from small to medium. Only two studies included in the present meta-analysis reported IQ scores separately for males and females (Offord, Reference Offord1974; Welham et al. Reference Welham, Scott, Williams, Najman, Bor, O'Callaghan and McGrath2009b). These studies indicated that, among the participants who subsequently developed schizophrenia, the males obtained lower IQ scores than the females. However, we did not observe a larger effect size in a study that examined only males (Osler et al. Reference Osler, Lawlor and Nordentoft2007) than in studies that included both males and females.
It is unclear whether lower than average IQ is an antecedent specific to schizophrenia. Lower than average IQ has been reported to characterize children (Van Os et al. Reference Van Os, Jones, Lewis, Wadsworth and Murray1997; Koenen et al. Reference Koenen, Moffitt, Roberts, Martin, Kubzansky, Harrington, Poulton and Caspi2009) and young adults, particularly males (Zammit et al. Reference Zammit, Allebeck, David, Dalman, Hemmingsson, Lundberg and Lewis2004; Mortensen et al. Reference Mortensen, Sørensen, Jensen, Reinisch and Mednick2005; Tiihonen et al. Reference Tiihonen, Haukka, Henriksson, Cannon, Kieseppa, Laaksonen, Sinivuo and Lonnqvist2005; Urfer-Parnas et al. Reference Urfer-Parnas, Lykke Mortensen, Sæbye and Parnas2009), who subsequently develop mental disorders other than schizophrenia. However, these findings are inconsistent and several studies have failed to identify IQ differences among children/adolescents who subsequently developed bipolar disorder (Cannon et al. Reference Cannon, Caspi, Moffitt, Harrington, Taylor, Murray and Poulton2002; Reichenberg et al. Reference Reichenberg, Weiser, Rabinowitz, Caspi, Schmeidler, Mark, Kaplan and Davidson2002; Zammit et al. Reference Zammit, Allebeck, David, Dalman, Hemmingsson, Lundberg and Lewis2004).
The present meta-analysis indicated that, by age 16, individuals who subsequently developed schizophrenia/SSD displayed poorer motor function than their peers who remained healthy. The lack of heterogeneity evidenced in this domain suggests that motor dysfunction is the most robust characteristic that distinguishes children/adolescents who subsequently develop schizophrenia. As noted previously, this is one of the most consistent findings in the literature on the antecedents of schizophrenia (Schenkel & Silverstein, Reference Schenkel and Silverstein2004). Also consistent with these findings are the results of three studies that could not be included in the present meta-analysis of motor function because of lack of available data (Crow et al. Reference Crow, Done and Sacker1995), or overlapping samples (Schiffman et al. Reference Schiffman, Walker, Ekstrom, Schulsinger, Sørensen and Mednick2004; Cannon et al. Reference Cannon, Moffitt, Caspi, Murray, Harrington and Poulton2006). These studies all reported deficits in motor function among children/adolescents aged 7–13 years who developed schizophrenia/SSD in adulthood. However, one study (Walker et al. Reference Walker, Savoie and Davis1994) observed a reduction in motor dysfunction with increasing age among children who later developed schizophrenia, which may reflect the insensitivity of measures of motor function or age-related improvements in motor function similar to those reported among typically developing children. In addition, one prospective study of a birth cohort that repeatedly assessed children did not detect motor function deficits at every age among those who later developed schizophreniform disorder (Cannon et al. Reference Cannon, Caspi, Moffitt, Harrington, Taylor, Murray and Poulton2002). Given the relative paucity of prospective longitudinal studies that have assessed motor function through childhood and adolescence with repeated assessments, it is unclear whether motor dysfunction is present across all periods of development among individuals developing schizophrenia/SSD.
Whether or not children and adolescents developing schizophrenia/SSD do poorly in school is currently unclear. Two reviews found that repeating a school year and achieving poor grades were associated with an increased risk of developing SSD (MacCabe, Reference MacCabe2008; Welham et al. Reference Welham, Isohanni, Jones and McGrath2009a). The results of the present meta-analyses indicated no significant difference in either overall academic achievement or performance on mathematics tests between individuals who did and did not later develop schizophrenia/SSD. This may be due, in part, to the inclusion of a study with a poorly matched comparison group (Ang & Tan, Reference Ang and Tan2004). In that study, the individuals who subsequently developed psychosis showed deterioration in mathematics test scores from age 12 to age 16. In the present meta-analysis, all of the participants in the studies that assessed performance in mathematics were aged 11 or 12 years. Furthermore, a study excluded from the Mathematics domain due to age of participants also found significant differences in mathematics achievement between individuals aged 12–18 years who did and did not develop schizophrenia in adulthood (Watt & Lubensky, Reference Watt and Lubensky1976). It is possible that individuals who later develop schizophrenia/SSD display a decline in performance on mathematics tests after age 12. Although significant heterogeneity was reported for both domains of academic performance, it could not be explored because of the limited number of studies meeting inclusion criteria. It is possible that the heterogeneity observed in the present meta-analysis and the inconsistent results across studies reflect differences in educational systems and measures of academic achievement that may preclude the examination of these domains in future meta-analyses.
Strengths and limitations
The present study is characterized by two principal strengths. One, the meta-analyses included only studies that had assessed performance in youth aged 16 years or younger. None of the 23 studies included in the meta-analyses reported that they had assessed prodromal symptoms at the same time as they assessed cognitive and/or motor performance. However, given the participants' age at the time of assessment, it is unlikely that the participants who subsequently developed schizophrenia/SSD had entered the prodromal phase of illness. Thus, the present results suggest that deficits in IQ and motor function emerge during childhood and early adolescence, prior to the onset of the prodrome. A second strength of the present meta-analyses was the examination of four domains of functioning: IQ, motor function, general academic achievement, and achievement in mathematics tests. Only the IQ domain had been examined previously using meta-analytic techniques. Despite using broad search terms to identify relevant studies, only 23 studies met final criteria for inclusion in the present meta-analyses. This was primarily due to the limited number of studies of cognitive and motor function in youth aged 16 years or younger who subsequently developed schizophrenia/SSD. More evidence is needed. Although the small number of studies precluded the examination of the heterogeneity of results obtained in domains other than IQ, the strict criterion requiring that participants had been assessed by age 16 allowed us further understanding of the development of schizophrenia/SSD.
A potential caveat relates to the use of meta-analytic methods for comparisons of cognitive and motor function among children/adolescents of differing ages, which may fail to reflect the discontinuous nature of cognitive development (Harris, Reference Harris1995). However, of the 23 studies included in the present meta-analyses, 18 assessed participants at 13 years or under, and only six studies examined participants with an age range of more than 3 years (Offord, Reference Offord1974; Watt & Lubensky, Reference Watt and Lubensky1976; Walker et al. Reference Walker, Savoie and Davis1994; Ott et al. Reference Ott, Spinelli, Rock, Roberts, Amminger and Erlenmeyer-Kimling1998; Amminger et al. Reference Amminger, Schlögelhofer, Lehner, Ott, Friedrich and Aschauer2000; Sørensen et al. Reference Sørensen, Mortensen, Schiffman, Ekstrøm, Denenney and Mednick2010). As only a few studies reported results separately for males and females, the meta-analyses could not contribute to the growing evidence on sex differences in the development of schizophrenia/SSD.
Conclusions
The meta-analyses provide evidence that among youth aged 16 years or younger, individuals who subsequently developed schizophrenia/SSD displayed lower IQ and poorer motor function than youth who did not develop illness. These results extend previous findings by showing that these deficits precede the onset of illness and of the prodrome. The results also endorse the view that schizophrenia, at least in part, represents a disorder of neurodevelopment. Stable cognitive and motor deficits in childhood and early adolescence are potential targets for interventions that may modulate illness development or reduce the extent of dysfunction present in individuals who develop schizophrenia/SSD.
Acknowledgements
We thank M. Cannon, A. Caspi, T. Crow, J. Done, L. Erlenmeyer-Kimling, R. Fuller, M. Isohanni, J. Jokelainen, P. Jones, J. MacCabe, T. Moffit and M. Osler for providing the data necessary to complete the meta-analyses, and D. Stahl for statistical advice.
H.D. was supported by a Ph.D. studentship from the NIHR BRC. K.R.L. was supported in part by an NIHR Career Development Fellowship.
Declaration of Interest
All authors are affiliated with the National Institute for Health Research (NIHR) Specialist Biomedical Research Centre (BRC) for Mental Health at the South London and Maudsley National Health Service (NHS) Foundation Trust and the Institute of Psychiatry, King's College London, UK.
