Neuromotor and attentional dysfunctions appear to be putative developmental precursors to schizophrenia. They consistently appear with increased frequency in high-risk, birth cohort, and follow-back studies. In a number of high-risk studies, infancy and childhood neuromotor impairments predicted the later onset of schizophrenia-related psychosis. Attentional impairments during middle childhood and early adolescence in the New York High-Risk project predicted the development of schizophrenia-related psychosis.

The endophenotypic indices that appear to have the greatest predictive validity are neuromotor dysfunction and impaired performance on measures that tap processing under high attention demands, or measures of secondary memory. It remains unclear whether these mea sures tap schizophrenic-related processes specifically. A number of the measures (including continuous performance tests, partial-report-span-of-apprehension tasks, and secondary verbal memory tests) that are sensitive to subtle neurocognitive impairments in children at risk for schizophrenia in middle childhood and adolescence also detect neurocognitive impairments in children with ADHD and learning disabilities. It is unlikely that these impairments have cross-sectional diagnostic specificity.

Although the ability of childhood and ado-lescent measures of attention to predict schizophrenia-related psychosis in the New York High-Risk project is promising, those results need to be replicated in an independent sample. Future studies will need to determine the extent to which childhood and adolescent neurocognitive measures predict schizophrenia-related psychosis conditioned on the presence of a second risk factor—having a parent who is schizophrenic. In effect, the analyses reported by the New York High-Risk project contained two risk factors that predicted schizophrenia-related psychosis: being the child of a schizophrenic parent and having attentional, verbal short-term memory or neuromotor impairments. These factors did not predict the onset of schizophrenia in the children of parents with an affective disorder nearly as well as they did in the children of parents with schizophrenia. As noted above, children with other, more common psychiatric diagnoses show deficits on these types of tasks. More research is needed on the diagnostic accuracy of these measures when they are used in the general pediatric population before they can be used to screen children for precursors for schizophrenia. At present, all that we know is that these measures have some promise in predicting which children who have a parent with schizophrenia are likely to develop a schizophrenic disorder themselves.

What is needed in the next generation of studies is not merely the demonstration of group mean differences between high-risk and control groups. If endophenotypic measures are to be used as candidates for preventive intervention programs, then diagnostic accuracy analyses that specify the sensitivity and specificity of tasks by means of various cutting scores need to be

developed. Cutting scores can be created, depending on the purpose, that optimize sensitivity (detecting true positives) or specificity (false negatives). For example, if the intervention can produce significant adverse events, it might be desirable to set a cutting score to minimize false positives.

Poor peer relations, a preference for solitary play, a “schizoid” social development, various nonpsychotic symptoms (particularly internalizing symptoms), and poor affective control occur frequently during middle childhood and early adolescence in high-risk, birth cohort, and follow-back studies. While these behavior problems and symptoms are precursors of schizophrenia, they are not diagnostically specific. Many of these symptoms are associated with other psychiatric disorders. For example, poor affective control is both a symptom of and precursor to affective disorders. Poor peer relationships are associated with the presence of both externalizing and internalizing disorders. There are relatively few data on the diagnostic accuracy (i.e., specificity and sensitivity) of symptoms and behavior problems detected in middle childhood and early adolescence as predictors of schizophrenia-related psychoses.

The behavior problems and symptoms that are putative precursors of schizophrenia are associated with psychiatric disorders (e.g., depression and ADHD) that are much more common than schizophrenia in the general population. Thus high false-positive rates may result if these symptoms are used in the general pediatric population in an attempt to identify individuals likely to develop schizophrenia.

There is great interest in developing preventive interventions for schizophrenia, in part because of the belief that once the disorder emerges, a neurodegenerative process is initiated that can only be partially forestalled by currently available treatments. The neurocognitive impairments, nonschizophrenic symptoms, and behavior problems that are putative developmental precursors of schizophrenia may have important implications for the development of preventive interventions for this disorder. These precursors could be used to identify children who might benefit from preventive intervention and serve as targets of interventions.

The neurocognitive impairments that are putative developmental precursors of schizophrenia have potential utility in identifying candidates for preventive interventions. Depending on the risk profile of the intervention, cutting scores on neurocognitive indices could be constructed to either maximize sensitivity or minimize false positives. However, as noted above, before the cutting scores for putative neurocognitive precursors of schizophrenia can be applied to the general pediatric population, additional research is required to evaluate the diagnostic efficiency of these measures in populations without a genetic risk. The neurocognitive precursors of schizophrenia seem to be unlikely targets for preventive interventions. There is no evidence that mitigating attentional, memory, and neuromotor impairments forestalls the development of schizophrenia-related psychoses. Identifying neurocognitive precursors of schizophrenia does advance attempts to develop new somatic treatments for schizophrenia by helping to elucidate the dysfunctional neural networks that underlie this complex disorder.

The diagnostic accuracy of the behavior problems and symptoms that are putative precursors of schizophrenia thus far identified in high-risk, birth cohort, and follow-back studies have not been carefully examined. Given the nonspecificity of these behavior problems and symptoms, it seems likely that they would yield high rates of false positives if used to identify candidates for preventive interventions for schizophrenia. It may be that clinical features more proximal to the onset of schizophrenia-related psychoses, such as prodromal signs and symptoms, have greater diagnostic accuracy in predicting which children will develop schizophrenia. A number of research groups are currently addressing this question.

The behavior problems and symptoms that are putative precursors of schizophrenia are potentially interesting targets for interventions. To the extent that poor peer relations, the presence of internalizing symptoms, and poor affective control pose difficulties for the child and parent,

they become worthy targets of therapeutic interventions. Behavioral (e.g., social skills training) and pharmacological (mood stabilizing drugs) interventions for these problems are based on symptomatic presentations. The nonspecificity of these problems is not particularly problematic in this case. While there is no reason to believe that successfully enhancing social skills and controlling affective symptoms will forestall the development of schizophrenia, there is good reason to believe that enhancing social skills and controlling affective symptoms will enhance the current quality of life and may also improve the postpsychotic episode adaptation. The best predictor of postpsychotic psychosocial functioning is the level of premorbid social competencies.

One of the best ways to develop early-intervention efforts for schizophrenia is to start by identifying key features of those individuals who are or will become schizophrenic and determine how these features differ from those seen in normal individuals who are not ill and are not likely to ever become afflicted with the illness. Several research designs can accomplish this goal. For example, cross-sectional studies of patients and control subjects can be used to characterize each group on as many potentially meaningful variables as possible, including behavior, personality, social activity, neuropsychological abilities, brain structure and function, and genetics. One problem with this method, however, is that any differences observed between the two groups cannot necessarily be attributed a causal role in the development of disease. For example, if total brain volume were lower among a group of schizophrenic patients than it was among a group of well-matched controls, this might indicate that low brain volume is a precursor or predictor of the development of schizophrenia. However, from such a cross-sectional design, it is unclear if the brain volume deficit in the patient group actually preceded the onset of schizophrenic illness. In fact, it is possible that it did, but it is also possible that the onset of schizophrenia caused a decline in brain volume due to some degenerative process. Alter natively, other factors, such as treatment with antipsychotic medication, may have precipitated the decline in brain volume. It is further possible that the brain volumetric decline in the patient group was concurrent with the onset of illness but causally unrelated to it.

Numerous cross-sectional studies have unearthed a wealth of information regarding the ways in which schizophrenic patients are different from patients with other psychiatric illnesses and from normal control subjects. However, because of the limitations on causal inference that exist in these types of studies, their results can only guide further research; they are not powerful enough to dictate a specific pattern of behavioral, neuropsychological, or biological characteristics that would be useful for identifying individuals for targeted prevention efforts. As already mentioned, studies of individuals with prodromal signs of schizophrenia and individuals with schizotaxia provide more insight into those traits that precede the disorder than do cross-sectional studies. Thus, great efforts have been made to enable identification of individuals in the earliest stages of the illness or even in the premorbid period so that they may be targeted for intervention.

By characterizing the prodromal phase of schizophrenia, subtle changes in behavior have been noted in those who are beginning to deteriorate into the early stages of the disease, and these changes are now being used to identify other clinically at-risk individuals for inclusion in early intervention programs. Some of the more pronounced changes observed during the prodrome occur in domains of thought, mood, behavior, and social functioning (Phillips, Yung, Yuen, Pantelis, & McGorry, 2002). Specifically, difficulties in concentration and memory may emerge, as well as preoccupations with odd ideas and increased levels of suspiciousness. Mood changes may include a lack of emotionality, rapid mood changes, and inappropriate moods. Beyond simply odd or unusual behavior, the prodrome may also be characterized by changes in sleep patterns and energy levels. Social changes can be quite marked, with withdrawal and isolation being the most predominant features. These characteristics may be particularly informative of the disease process in schizophrenia, be

cause they are by definition not related to the effects of medication or the degenerative effects of being ill for a prolonged period.

Perhaps the most powerful window into the premorbid changes in preschizophrenic individuals comes from the longitudinal study of children and adolescents who are genetically at high risk for the illness. By studying the biological children of schizophrenic parents, the clinical, behavioral, and biological features of schizotaxia can be revealed. Longitudinal studies of individuals such as these, who harbor the latent genetic liability toward schizophrenia, can be extremely informative for early intervention and prevention efforts because they can track the emergence of schizophrenia precursors before any signs of illness are apparent. Thus, any differences observed between children of schizophrenic patients and children of control subjects can be definitively attributed to factors other than the effects of antipsychotic medication, the degenerative effects of the illness, or any other factors that are subsequent to disease onset. The observed differences can be viewed as antecedents to the illness, which is as close to a causal relationship as can be ascribed in human research studies in which group membership cannot be experimentally assigned.

Studies of children of patients with schizophrenia have yielded a variety of findings of altered behavioral, neuropsychological, and biological processes. The richness and diversity of measures taken on these subjects can make profiling the premorbid genetic susceptibility to schizophrenia difficult. Such studies have also produced some surprisingly uniform findings, which simplify our understanding of what may be the most central or universal deficits among those who are at the highest risk for schizophrenia.

Certain personality characteristics seem to reliably differentiate children of schizophrenic parents from children of control subjects (Miller et al., 2002). For example, schizotypal personality features, including social withdrawal, psychotic symptoms, socioemotional dysfunction, and odd behavior, have been shown to precede the onset of psychosis among genetically high-risk children. Deficits of social functioning are also commonly observed in this group (Dworkin et al., 1993). Specifically, children of schizophrenic patients are more likely than children of controls to have more restricted interests, significantly poorer social competence (especially in peer relationships and hobbies and interests), and greater affective flattening. Some neuropsychological deficits have also been reliably observed in these high-risk individuals (Asarnow & Goldstein, 1986; Cosway et al., 2000; Erlenmeyer-Kimling & Cornblatt, 1992; Schreiber et al., 1992). For example, several studies have replicated a pattern of impaired discrimination, sustained attention, and information processing on the visual continuous performance test among children of schizophrenic patients. These high-risk individuals also exhibit marked impairments on memory for verbal stimuli and in executive functioning, as well as neuromotor deficits such as soft neurological signs, gross and fine motor impairments, and perceptual-motor delays.

Perhaps underlying these personality, social, and neuropsychological deficits, children of schizophrenic patients have also been shown to have altered brain structure and function compared to that of children of control subjects (Berman, Torrey, Daniel, & Weinberger, 1992; Cannon et al., 1993; Liddle, Spence, & Sharma, 1995; Mednick, Parnas, & Schulsinger, 1987; Reveley, Reveley, & Clifford, 1982; Seidman et al., 1997; Weinberger, DeLisi, Neophytides, & Wyatt, 1981). The most commonly observed structural brain abnormality among children of schizophrenic patients is a reduced volume of the hippocampus and amygdala region. Loss of volume in the thalamus has also been observed in these children, and there has been some support for enlarged third ventricular volume and smaller overall brain volume in this group. Children of schizophrenic patients also have been found to exhibit linear increases in cortical and ventricular cerebrospinal fluid to brain ratios with increasing genetic load—that is, children with the greatest number of affected biological relatives showed the highest ratios.

Ultimately, these clinical, behavioral, social, and biological profiles of risk for emergent schizophrenia will be augmented by information on specific genes that increase susceptibility. Recently, genes coding for neuregulin 1

(NRG1; Stefansson et al., 2002), nitric oxide synthase (NOS1; Shinkai, Ohmori, Hori, & Nakamura, 2002), and dystrobrevin-binding protein 1 (DTNBP1; Straub, Jiang, et al., 2002) have been reported to have an association with schizophrenia, but these findings will require verification. Many other polymorphisms have shown a positive association with the disorder, but attempts to replicate these findings have often failed. For several of these widely studied polymorphisms, meta-analysis has been used to clarify the presence or absence of a true allelic association with the disorder in the presence of ambiguity. In fact, using this approach, some candidate genes, including those that code for the serotonin 2A receptor (HTR2A) and the dopamine D2 (DRD2) and D3 (DRD3) receptors, have already been shown to have a small, but reliable, association with the disorder (Dubertret et al., 1998; Glatt, Faraone, & Tsuang, 2003; Williams, McGuffin, Nothen, & Owen, 1997). Eventually, other gene variants, including perhaps NRG1, NOS1, and DTNBP1, will be found to be reliably associated with schizophrenia. This may make it possible to create a genetic risk profile that will be predictive of future onsets of schizophrenia, especially in combination with other known risk indicators.

Together, the various abnormal features of children of schizophrenic patients provide a “composite sketch” of the underlying premorbid susceptibility toward schizophrenia. Because the probability of developing schizophrenia among children of one or two affected individuals (12% and 46%, respectively) is far greater than that probability among children of control subjects (1%), these abnormalities signal the subsequent development of schizophrenia with a relatively high degree of sensitivity and reliability. However, it is also clear that these trends are not absolute, and many children of schizophrenic patients will not exhibit these signs, nor will they ever develop schizophrenia.

It has been recognized for some time that the duration of untreated illness in schizophrenia is correlated with the prognosis for the disease, such that those with the longest period of untreated psychosis experience the least favorable outcomes (Browne et al., 2000). Recently, it has also been discovered that outcome correlates with the duration of illness as measured from the onset of the prodrome rather than only from the onset of frank psychosis. From this line of evidence, the rationale for early-intervention efforts was born. It was reasoned that, if early treatment of the illness led to a more favorable outcome, early intervention even before the onset of the illness might further inhibit the progression of the illness, either delaying its onset, decreasing its severity, or both.

A fundamental question in designing early intervention protocols is, what will be the target of the intervention? There is no single best answer to this question, which may be why various targets are being used in current early intervention efforts. The earliest interventions might realize the greatest opportunities to divert high-risk individuals from the subsequent development of schizophrenia, but the ability to predict schizophrenia accurately might be greatest in the period closest to disease onset. For example, targeting attention problems in young children of schizophrenic parents might allow the identification of the children who are at highest risk of transitioning to psychosis and afford ample time to intervene in that process; yet because of the restricted sensitivity and specificity of this deficit, targeting attention problems may also cause some high-risk children to be excluded from the protocol while, inevitably, some of the children who were included in the protocol would not go on to develop the illness. Targeting the changes of the prodrome, by contrast, such as the emergence of odd behaviors or increased suspiciousness, might lower false-positive and false-negative classification errors, but the ability of the intervention protocol to influence the course of the illness might be relatively restricted compared to earlier interventions. Thus, a balance must be maintained between the potential effectiveness of the intervention and the specificity of the intervention to the target population.

Another key question in developing early intervention protocols is, at what level should the intervention be administered? Again, this is a