The 22q11.2 deletion syndrome, also known as velocardiofacial syndrome or DiGeorge syndrome, is the most common microdeletion genetic disorder known in human beings, occurring in at least 1:5,000 live births (1). It is caused by a microdeletion in the long arm of chromosome 22 and is associated with congenital malformations and cognitive deficits (1, 2). The 22q11.2 deletion syndrome is an appealing model for studying risk factors for the emergence of schizophrenia and other psychotic disorders, as one-third of all individuals with 22q11.2 deletion syndrome develop schizophrenia-like psychotic disorders (3, 4).

From early childhood, abnormal behaviors and an increased rate of psychiatric disorders are already present in subjects with 22q11.2 deletion syndrome. While the schizophrenia-related psychotic disorders usually evolve only during late adolescence or early adulthood, children and adolescents with 22q11.2 deletion syndrome have a high frequency of nonpsychotic psychiatric disorders, including attention deficit hyperactivity disorder (ADHD), oppositional defiant disorder, anxiety and affective disorders, autism, and obsessive-compulsive disorder (OCD) (5–7). Some of these early abnormal neuropsychiatric symptoms could be premorbid signs for the later development of psychosis (8, 9). Cross-sectional studies of children with 22q11.2 deletion syndrome indicate that one-third to one-half of affected individuals demonstrate subsyndromal evidence of definite or suspected signs and symptoms of psychosis (5, 6).

The purpose of this study was twofold. First, we sought to evaluate the psychiatric, behavioral, and adaptive developmental trajectories of individuals with 22q11.2 deletion syndrome in comparison with individuals with idiopathic developmental disability during the critical period from childhood and early adolescence to late adolescence and early adulthood. We predicted that significantly more individuals in the 22q11.2 deletion syndrome group would develop psychotic disorders than in the comparison group. Our second objective was to identify the profile of early risk factors for the development of psychotic disorders in 22q11.2 deletion syndrome. On the basis of our previous findings (3), we assumed that the catechol O-methyltransferase (COMT) low-activity allele (COMT Met) would predict the development of psychotic disorders by the time of the follow-up evaluation.

On the basis of findings from the neuroimaging literature on schizophrenia in the general population (10, 11) and on 22q11.2 deletion syndrome (12, 13), we also hypothesized that abnormal prefrontal and temporal gray matter volumes at baseline would be associated with the development of psychotic disorders. Finally, we predicted that subjects with prodromal psychotic symptoms at the time of the first evaluation would be more likely to have developed a psychotic disorder at follow-up.

Subjects

The baseline assessment, conducted between 1998 and 2000, included 31 children with 22q11.2 deletion syndrome and a comparison group of 29 children with idiopathic developmental disability matched for age, gender, and IQ. Recruitment and sample characteristics have been described elsewhere (6). The 22q11.2 deletion syndrome group included only children in whom the microdeletion was confirmed with a fluorescence in situ hybridization test. Similarly, absence of the 22q11 deletion was confirmed in all comparison subjects.

A follow-up assessment of the initial sample, conducted between 2003 and 2005, included 28 children from the 22q11.2 deletion syndrome group (90.3%) and 23 from the developmental disability group (79.3%). Dropouts were due to the inability to locate subjects whose family had left the area (two from the 22q11.2 deletion syndrome group and four from the developmental disability group) or to subjects’ choosing not to participate in the follow-up evaluation (one from the 22q11.2 deletion syndrome group and two from the developmental disability group). In both groups, subjects who dropped out of the study were similar to their counterparts who participated in the follow-up evaluation in baseline age and severity of psychiatric symptoms (as measured by the Child Behavior Checklist—Parent Version [CBCL; 14]). Full-scale IQ scores tended to be lower in subjects who dropped out compared with those who participated at follow-up (57.0 [SD=17.4] versus 70.5 [SD=16.1] in the 22q11.2 deletion syndrome group and 68.3 [SD=21.8] versus 74.4 [SD=23.5] in the developmental disability group).

There were no significant differences between the groups at follow-up in age at baseline assessment (mean=12.5 years [SD=3.9] and mean=12.9 years [SD=3.3] for the 22q11.2 deletion syndrome and developmental disability groups, respectively), age at follow-up assessment (mean=17.4 years [SD=3.7] and mean=18.2 years [SD=3.8], respectively), gender distribution (males:females, 18:10 and 14:9, respectively), full-scale IQ (mean=70.5 [SD=16.1] and mean=74.3 [SD=23.6], respectively), and duration of parents’ education (mean=16.4 years [SD=2.2] and mean=17.1 years [SD=3.0], respectively). The mean time between assessments was shorter for the 22q11.2 deletion syndrome group than for the developmental disability group (mean=4.8 years [SD=0.8] versus mean=6.0 years [SD=2.6], respectively).

After providing a complete description of the study to the subjects and their parents, written informed consent was obtained at both assessments, under protocols approved by the institutional review board of Stanford University.

Genotype

Blood samples were drawn from the 22q11.2 deletion syndrome group to determine genotype. The COMT Val108/158Met polymorphism (rs165688) was genotyped by polymerase chain reaction and restriction digestion according to standard techniques and was determined by NlaIII digestion (NlaIII+=Met, NlaIII–=Val) (15).

Neuropsychiatric, Adaptive Functioning, and Cognitive Assessments

Interviewers at follow-up were blind to baseline evaluations and diagnoses. They were also blind to group status, although a minority of subjects in the 22q11.2 deletion syndrome group exhibited clinical features of the syndrome, such as dysmorphic facies and hypernasal speech. The baseline psychiatric evaluation included the following:

1. The Child Behavior Checklist—Parent Version. This instrument is a well-established and widely used questionnaire, both for typically developing children (14) and for children with developmental disabilities (16). The scale yields a total problem score and eight syndrome scores.

2. The computerized Diagnostic Inventory for Children and Adolescents (17), a DSM-IV-based structured interview that was administered by a child and adolescent psychiatrist to participants’ parents.

3. The screening-question portion of the parent version of the Schedule for Affective Disorders and Schizophrenia for School-Aged Children—Present and Lifetime Version (K-SADS-PL) (18). If any screening item was positive, the full K-SADS-PL section on the relevant possible diagnosis was administered. In addition, a 3-point psychosis scale was created on the basis of the K-SADS-PL psychosis scores. A score of 0 indicated no hallucinations or delusions; 1 indicated subthreshold (i.e., suspected or likely) hallucinations or delusions; and 2 indicated definite hallucinations or delusions. Each subject’s diagnostic information was reviewed by at least two of five child psychiatrists (D.G., C.F., H.K., S.E., and A.L.R.), and a consensus diagnosis based on DSM-IV criteria was made.

The follow-up psychiatric evaluations were similar to the baseline evaluations but included some modifications because some of the subjects were young adults at follow-up. Thus, for subjects 19 years of age and older, we used the Adult Behavior Checklist (19), an adult version of the CBCL, also completed by parents. For the longitudinal analyses, we used six symptom scales and the internalizing and externalizing group scores, which are similar in the child and adult questionnaires. In addition, subjects over age 18 were evaluated with the Structured Clinical Interview for DSM-IV Axis I Disorders (SCID) (20). All subjects were interviewed at follow-up by a child and adolescent psychiatrist, who completed the Brief Psychiatric Rating Scale (BPRS) based on his or her evaluation of the subject (21). The interview was extended to include the K-SADS-PL or SCID sections assessing psychosis. The interviewer also asked subjects to verify uncertain symptoms suggested by their parents (e.g., depressive, compulsive, or phobic symptoms).

Adaptive behavior was assessed at both baseline and follow-up with the Vineland Adaptive Behavior Scales—Interview Edition (22), which were completed on the basis of interviews with the parents. Standard scores were calculated on the basis of the age-appropriate national standards up to the age of 19 years (22). For subjects above age 19, the norms of the upper age level were used (18 years and 8 months to 19 years), as suggested by the manual (22).

For cognitive assessment, the Wechsler Intelligence Scale for Children, 3rd edition (23), was used for subjects age 17 and younger, and the Wechsler Adult Intelligence Scale, 3rd edition (24), for those above age 17.

Magnetic Resonance Imaging Protocol

The 22q11.2 deletion syndrome group underwent magnetic resonance imaging (MRI) scans. Coronal images were acquired with a GE 1.5-T scanner (General Electric, Milwaukee) using a three-dimensional volumetric radiofrequency spoiled gradient echo pulse sequence with the following scan parameters: TR=35 msec, TE=6 msec, flip angle=45°, number of excitations=1, matrix size=256×192, field of view=24 cm², slice thickness=1.5 mm, 124 contiguous slices. MRI scans were imported into BrainImage (Stanford University, Stanford) for semiautomated whole-brain segmentation and quantification using previously described and validated methods (25) that produce volume measurements of gray matter, white matter, CSF, and total tissue for the whole brain and for the four cerebral lobes. The prefrontal cortex was defined as all frontal cortical gray matter lying anterior to a coronal plane intersecting the most anterior point of the genu of the corpus callosum (3, 26).

Data Analysis

Pearson chi-square tests were used to assess the difference between groups in the frequency of psychiatric disorders at baseline. For between-groups comparisons of the change in frequency of psychiatric disorders from baseline to follow-up, logistic regressions were done for each psychiatric diagnosis. In each analysis, the diagnostic status at follow-up was the outcome variable, and baseline psychiatric diagnosis by group interaction was predictor. Unpaired t tests were used for between-groups comparison of baseline CBCL and Vineland Adaptive Behavior Scales subscale scores. Longitudinal change in CBCL and Vineland subscale scores was assessed by repeated-measures analyses of variance (ANOVAs) with group as the between-subjects factor and CBCL and Vineland subscale scores at baseline and follow-up as the within-subject factor. To clarify the interactions, we calculated the effect sizes (standardized mean differences) within each group. To further investigate significant interactions, we conducted post hoc pairwise t tests separately for the two groups. As the repeated-measures ANOVA was used 13 times, the significance threshold was corrected to an alpha of 0.004. For the longitudinal prediction of follow-up psychotic disorders and BPRS scores in 22q11.2 deletion syndrome, we used hierarchical/stepwise logistic and linear regressions.

Longitudinal Change in Psychiatric Morbidity

There were no significant differences between the 22q11.2 deletion syndrome group and the developmental disability group at baseline in the frequency of any psychiatric disorder, and none of the participants had a psychotic disorder (Table 1). At follow-up, significantly more subjects with 22q11.2 deletion syndrome than comparison subjects fulfilled DSM-IV criteria for a psychotic disorder (p=0.01). The psychotic disorders in the 22q11.2 deletion syndrome group were schizophrenia or schizoaffective disorder (N=6), schizophreniform disorder (N=2), and psychotic depression (N=1). One subject in the developmental disability group met criteria for psychotic depression. The 22q11.2 deletion syndrome group had a higher mean BPRS score at follow-up (35.7 [SD=12.5]) compared with the developmental disability group (28.0 [SD=7.1], t=2.7, df=49, p=0.01).

The logistic regression showed that there was a significant effect of anxiety disorder status (presence or absence) at baseline by group interaction on anxiety disorder status at follow-up (p<0.05). This difference in anxiety disorders at follow-up was mainly a result of the frequency of anxiety disorders decreasing in the developmental disability group and slightly increasing in the 22q11.2 deletion syndrome group. No other significant differences were observed.

Within-Subject Change in CBCL and Vineland Scores

At baseline, the CBCL externalizing scores were significantly higher in the comparison group than in the 22q11.2 deletion syndrome group (Table 2). No other significant differences were found between groups on any other CBCL or Vineland score. Repeated-measures ANOVAs showed that, after correction for multiple comparisons, changes in total problem, externalizing, and aggressive behavior scores differed significantly between groups, with the comparison group exhibiting a greater decrease in scores from baseline to follow-up compared with the 22q11.2 deletion syndrome group (Table 2). Post hoc pairwise t tests indicated that total problem (t=4.5, df=22, p<0.0001), externalizing (t=2.6, df=22, p<0.05), and aggressive behavior (t=3.5, df=22, p<0.005) scores significantly declined in the comparison group and did not significantly change in the 22q11.2 deletion syndrome group.

The CBCL thought problems scores for the 22q11.2 deletion syndrome group remained stable, although a large proportion of subjects in this group developed psychotic disorders; those with psychotic disorders had higher thought problems scores at follow-up than those without psychotic disorders (70.2 [SD=6.5] versus 61.6 [SD=8.6]; t=–2.7, df=26, p=0.01). In the comparison group, thought problem scores significantly declined between baseline and follow-up (t=2.4, df=21, p<0.05).

Prediction of Psychosis in 22q11.2 Deletion Syndrome

To identify the best set of baseline predictors for the development of psychotic disorders at follow-up, forward hierarchical/stepwise regression analysis was performed for the 22q11.2 deletion syndrome group. Because we found in a previous study that the COMT Met/Val genotype predicted the severity of psychotic symptoms (3), we entered the COMT genotype in the first step. Since our a priori hypothesis was that the presence of subthreshold and above-threshold psychotic symptoms at baseline would predict the development of a full-blown psychotic disorder at follow-up, the baseline psychosis scale was entered in the second step. At the third step, we entered all other predictors, including baseline prefrontal and temporal gray matter volumes adjusted for total cranial gray matter volume, baseline verbal IQ, and baseline CBCL factor scores on the withdrawn, anxious/depressed, thought problems, attention problems, and aggressive behavior subscales. The CBCL predictors were chosen on the basis of reports in the schizophrenia literature associating these prodromal or premorbid symptoms as potential risk factors for the development of psychosis (8, 9, 27–29).

Results from the logistic regression analysis indicated that at the first step, the COMT genotype was significantly associated with the presence of a psychotic disorder at follow-up (β=1.81, p<0.05). At the second step, psychotic symptoms at baseline were significantly associated with the presence of a psychotic disorder at follow-up (β=1.79, p<0.05). Of the third-step variables, only the CBCL anxious/depressed subscale scores were significantly associated with the presence of a psychotic disorder at follow-up (β=0.20, p<0.05).

Converging results were obtained when predicting follow-up BPRS scores with the same dependent variables. The hierarchical/stepwise linear regression analysis showed that COMT genotype predicted 24% of the variance in BPRS scores (β=0.49, r² change=0.24, p=0.01), baseline psychotic symptoms scores predicted 20% of the variance in BPRS scores (β=0.45, r² change=0.20, p<0.01), and CBCL baseline anxious/depressed scores accounted for an additional 23% of the BPRS scores’ variance (β=0.66, r² change=0.23, p=0.001). In addition, lower verbal IQ scores at baseline were significantly associated with BPRS scores and predicted an additional 7% of the variance (β=–0.26, r² change=0.07, p<0.05).

To evaluate possible interaction between baseline variables that significantly predicted the onset of psychotic disorders at follow-up, we conducted an additional linear regression analysis in which follow-up BPRS was the outcome variable and the predictors included the following interactions: COMT genotype by baseline psychotic symptoms, COMT genotype by baseline CBCL anxious/depressed scores, and baseline psychotic symptoms by baseline CBCL anxious/depressed scores. Baseline psychotic symptoms by baseline anxious/depressed scores strongly predicted BPRS scores (β=0.12, p=0.001); COMT genotype by baseline psychotic symptoms scores also strongly predicted BPRS scores (β=6.51, p=0.002). Together, these two predictors accounted for 61% of the variance in BPRS scores.

The CBCL subscales for withdrawn, thought problems, attention problems, and aggressive behavior as well as the baseline prefrontal and temporal gray matter volumes were not found to predict the development of psychotic disorders in 22q11.2 deletion syndrome in both regression analyses.

The association between psychotic symptoms at baseline and the presence of psychotic disorders at follow-up was specific to the 22q11.2 deletion syndrome group; none of the four subjects in the developmental disability group with subthreshold or above-threshold psychotic symptoms at baseline developed a psychotic disorder, whereas five of eight children in 22q11.2 deletion syndrome group who had psychotic symptoms at baseline had developed a psychotic disorder by follow-up.

The presence of any anxiety disorder (not including phobias) at baseline was strongly associated with the development of psychotic symptoms at follow-up (p<0.005). Interestingly, all four subjects in the 22q11.2 deletion syndrome group who had a diagnosis of OCD at baseline met diagnostic criteria for a psychotic disorder at follow-up.

To our knowledge, this is the first longitudinal study of psychiatric symptoms in individuals with 22q11.2 deletion syndrome. Our results indicate that children with 22q11.2 deletion syndrome exhibit more severe psychiatric symptoms than children with idiopathic developmental disability as they pass through adolescence. This finding was predominantly associated with the development of psychotic disorders in 32% of the 22q11.2 deletion syndrome sample. In addition, whereas the developmental disability group showed significant reductions in several CBCL subscale scores from baseline to follow-up, these scores remained high in the 22q11.2 deletion syndrome group.

The high rate of psychotic disorders in our sample of individuals with 22q11.2 deletion syndrome, whom we first assessed during late childhood or early adolescence and then reassessed during late adolescence or early adulthood, is in line with that reported for adults with 22q11.2 deletion syndrome (4). The mean age at onset of psychotic disorders in 22q11.2 deletion syndrome has varied greatly among studies, with a range of 11 to 26 years (4, 30, 31). This variability could be due in part to differences in the age of the samples. Another explanation, supported by the results of our longitudinal study, is that the evolution of psychotic disorders in 22q11.2 deletion syndrome is a gradual and protracted process that is manifested in childhood by subclinical psychotic symptoms and a predisposition for internalizing symptoms and disorders.

In a previous publication on this cohort (3), we reported that a genetic variation of the COMT gene is a risk factor for aberrant reduction in prefrontal gray matter and verbal IQ and for the development of psychotic symptoms. The present study shows that the presence of psychotic symptoms at baseline interacts with both the COMT genotype and the presence of baseline symptoms of anxiety or depression to dramatically increase the rate of development of psychotic symptoms by follow-up. These two interactions explained 61% of the variance in follow-up BPRS scores.

Occasional psychotic symptoms, such as hallucinations, are commonly reported by typically developing children (32, 33), and it is unclear whether these symptoms are associated with a greater risk of developing a psychotic disorder in young adulthood. Our findings indicate that for individuals with 22q11.2 deletion syndrome, there is a continuity and gradual exacerbation of psychotic symptoms from childhood to young adulthood.

The other major predictor for the development of psychotic disorders and symptoms in our 22q11.2 deletion syndrome subjects was a high baseline score on the CBCL anxious/depressed subscale. This high level of premorbid symptoms of anxiety and depression is consonant with findings reported in longitudinal risk studies of offspring of individuals with schizophrenia (9, 34). Anxiety disorders, especially OCD, are often reported to be comorbid with schizophrenia, both in the general population (35) and in individuals with 22q11.2 deletion syndrome (7). Our study shows that anxiety disorders, and OCD in particular, are robust early childhood indicators for the later development of psychosis in 22q11.2 deletion syndrome. Remarkably, all subjects with OCD at baseline had developed a psychotic disorder at follow-up. This may imply that specific genetic factors predispose individuals with 22q11.2 deletion syndrome to a developmental expression of the psychiatric phenotype that begins with obsessive-compulsive symptoms and later develops into psychosis.

The results of the linear regression suggest that lower verbal IQ in childhood is a risk factor in 22q11.2 deletion syndrome for the later development of psychotic disorders. This finding expands on our previous finding that a decline in verbal IQ is strongly associated with the emergence of psychotic disorders (3). Similarly, childhood schizophrenia in subjects who do not have 22q11.2 deletion syndrome is associated with lower IQ scores (36), and recent longitudinal studies indicate that declines in IQ occur several years before the emergence of schizophrenia (37).

Although ADHD is the most common psychiatric disorder in 22q11.2 deletion syndrome, occurring in 40%–50% of subjects, the presence of ADHD was not associated with a greater risk of the later development of psychosis in our sample. Adjusted gray matter volumes of the prefrontal and temporal cortices at baseline were not found to be risk factors for the development of psychotic disorders after other key variables were entered into the analysis. Further studies with larger samples are needed to detect additional predictors for psychosis, such as brain morphology, that our study did not have the power to assess because of its small sample size.

The findings of this longitudinal study have important clinical implications in the possible prevention of psychotic disorders in 22q11.2 deletion syndrome. Our results suggest that children with 22q11.2 deletion syndrome should be carefully and routinely screened for early signs of mild psychotic manifestations and internalizing symptoms. Because these high-risk symptoms are usually not disruptive—as opposed, for example, to hyperactivity and oppositional symptoms—they could be easily overlooked and go untreated. Preliminary data suggest that atypical antipsychotics delay or prevent the progression of prodromal symptoms to schizophrenia in adolescents and young adults (38–40). Thus, strong consideration should be given to the use of antipsychotic medication in children with 22q11.2 deletion syndrome who exhibit even mild (i.e., subsyndromal) psychotic symptoms. This treatment may prevent, or at least reduce the severity of, a subsequent full-blown psychotic disorder. Future research should address the potential for preventing or reducing the severity of psychosis by treating internalizing symptoms in children with this disorder.

Received March 29, 2006; revisions received July 16 and Sept. 26, 2006; accepted Oct. 12, 2006. From the Behavioral Neurogenetics Center, Schneider Children’s Medical Center of Israel, Petah Tiqwa and Sackler Faculty of Medicine, Tel Aviv University, Ramat Aviv, Israel; Center for Interdisciplinary Brain Sciences Research, Stanford University School of Medicine, Stanford; Department of Psychiatry and Behavioral Sciences, University of Washington, Seattle; and Department of Psychiatry, University of Geneva School of Medicine, Geneva. Address correspondence and reprint requests to Dr. Gothelf, Department of Child Psychiatry, Schneider Children’s Medical Center of Israel, 14 Kaplan St., Petah Tiqwa, Israel 49202; gothelf@post.tau.ac.il (e-mail).

All authors report no competing interests.

Supported by a NARSAD Young Investigator Award to Dr. Gothelf and by NIH grants HD31715, MH50047, and MH19908 to Dr. Reiss. The authors thank Drs. Helena C. Kraemer and Booil Jo for statistical consultation and Noopur Jain and Ira Patnaik for data processing.