Clinical and cognitive risk factors for psychotic symptoms in 22q11.2 deletion syndrome: a transversal and longitudinal approach

This cohort has been the subject of previous publications and the global research strategy has already been published elsewhere [e.g. 17]. Briefly, all participants were recruited through advertisements in patient association newsletters and through word of mouth. Written informed consent was obtained for all participants and their parents under the protocols approved by the Institutional Review Board of the Department of Psychiatry of the University of Geneva Medical School. The presence of a 22q11.2 microdeletion was confirmed using quantitative fluorescent polymerase chain reaction (QF-PCR).

The results from the present study were obtained by analysing cross-sectional data in an original sample of 104 individuals with 22q11DS. They were further completed by additional analyses performed on a cross-sectional subsample of 59 individuals and a longitudinal subsample of 56 individuals (see Table 1).

The original sample of 104 participants with 22q11DS of age 6–44 years [m _age = 14.83, SD = 7.53; 57 (55 %) females] was used to characterize the prevalence of hallucinations and delusional ideas. The sample was subdivided into three age groups to compare the prevalence of these manifestations across their lifespan: children aged 6–11 years (N = 45); adolescents aged 12–17 years (N = 39); and adults above 18 years (N = 20). Seventeen (16.3 %) participants were receiving psychotropic medication at the time of testing: six were on methylphenidate, six on antipsychotics, five on antidepressants, three on antiepileptic drugs and two on anxiolytic medication.

As data collected in this large sample remained relatively general, we performed more detailed analyses in a cross-sectional subsample of participants using the Structured Interview for Prodromal Syndromes (SIPS; see “Materials”). The SIPS was introduced later into the Geneva longitudinal study research protocol and is not appropriate for use with young children. SIPS data were available for a subsample of 59 participants (m _age = 16.93, SD = 5.67; 33 (56 %) females]. Patients assessed with the SIPS were significantly older compared with patients who were not assessed with the SIPS (t = −3.423, p = 0.001), but did not differ regarding gender distribution (χ² = 0.070, df = 1, p = 0.792), mean full-scale IQ (t = −0.515, p = 0.608) and mean BPRS positive symptom score (t = −1.460, p = 0.147). The prevalence of patients receiving psychotropic medication did not significantly differ between the two groups (χ² = 0.527, df = 1, p = 0.468). The difference in age was due to the fact that the SIPS was not administered to younger children.

Subsequent to cross-sectional analyses, we analysed the evolution of psychotic symptoms in our cohort using the longitudinal data collected to date with the Brief Psychiatric Rating Scale (BPRS; see “Materials”). These analyses enabled examining the developmental trajectories of positive psychotic symptoms as well as the association between positive psychotic symptoms and cognitive functioning. Longitudinal assessments were available for 63 participants, of whom we excluded 7 aged below 12 years at T2 evaluation to limit the probability of false negatives (i.e. participants with no apparent psychotic symptoms who may still develop symptoms later in life). The final longitudinal subsample was therefore composed of 56 patients [m _age at T1 = 16.60, SD = 6.73; 33 (59 %) females], with a mean interval of 3.86 years between the two visits (SD = 0.69 year, range 2.57–6.00 years). Patients for whom longitudinal data were not available did not differ from patients assessed longitudinally regarding age (t = −1.276, p = 0.205), gender distribution (χ² = 0.657, df = 1, p = 0.417), full-scale IQ (t = −0.924, p = 0.358) and mean BPRS positive symptom score (t = −0.921, p = 0.359). The prevalence of patients receiving psychotropic medication did not significantly differ between the two groups (χ² = 2.398, df = 1, p = 0.121).

The first analyses were based on cross-sectional data and performed using SPSS version 19. We first compared the prevalence of schizophrenia spectrum disorders (based on the DICA, SCID-I and K-SADS-PL), definite hallucinations and delusions, and subthreshold hallucinations and delusions (based on the K-SADS-PL) across age groups using Chi-square tests. Hallucinations were then characterized according to their modalities of expression (auditory, visual or combined forms); delusions were characterized according to their thematic forms (referential, persecutory or combined forms).

The second analyses were based on longitudinal data available for a subsample of 56 patients. We first examined the prevalence of different clinical trajectories of positive psychotic symptoms over time: absence of symptoms (i.e. score <3 on the three BPRS items assessing positive symptoms at T1 and T2 evaluation), transient symptoms (i.e. score ≥3 on at least one positive BPRS item at T1 and score <3 on positive BPRS items at T2), emerging symptoms (i.e. score <3 on positive BPRS items at T1 and score ≥3 on at least one positive BPRS item at T2) and enduring symptoms (i.e. score ≥3 on at least one positive BPRS item at T1 and T2).

We then examined the developmental trajectories of cognitive functioning in patients with and without positive psychotic symptoms at T2. A score ≥3 on at least one positive BPRS item was used to determine the presence of positive psychotic symptoms. These analyses were performed using mixed model regression analyses, allowing to model the within-subject factor as a nested variable [29]. Such mixed model analyses allow increased statistical power in longitudinal dataset such as ours, with variable time between visits and a large age range at T1 evaluation. Different random intercept models (constant, linear, quadratic and cubic) were proposed, using the nlmefit function in MATLAB R2011b (MathWorks). We used a Bayesian information criterion (BIC)-based model selection method, as the BIC-based methods are one of the most powerful model selection methods for the mixed models [30]. Then, a likelihood ratio test was used to quantify significant between-group differences in the cognitive trajectories over time. At the end of the analysis, two types of differences can be observed between the groups: shape differences (the trajectories of cognitive functioning in the two groups have a different shape) and intercept differences (the trajectories in the two groups have the same shape (i.e. they are parallel), but have different intercept values).

In the present sample, we observed a high prevalence of schizophrenia spectrum disorders in adults with 22q11DS, similar to what was typically observed in other research groups [e.g. 1]. Furthermore, we observed that the rate of schizophrenia spectrum disorders was also increased in the adolescent group (8 %), suggesting that early-onset psychosis in frequent in 22q11DS and may represent a clinical characteristic of the syndrome [31, 32]. Indeed, the onset occurred before the age of 18 years in 30 % of individuals with a schizophrenia spectrum disorder. Although the exact prevalence of early-onset schizophrenia (i.e. onset before 18 years) in schizophrenic patients without 22q11DS remains unknown, the 30 % prevalence observed in the present study is much higher than the estimated rate of early-onset schizophrenia [e.g. 33]. For this reason, child and adolescent psychiatrists unfamiliar with 22q11DS may misdiagnose psychotic symptoms in these adolescents. Because the duration of untreated psychosis is known to affect outcome [34], it thus becomes increasingly important to heighten practitioners’ awareness about the specificities of schizophrenia spectrum disorders in the context of 22q11DS.

When psychotic symptoms were examined in the same sample of patients, it appeared that a substantial proportion of them reported psychotic manifestations in the absence of a full-blown psychotic disorder and even at a very young age [see also 3, 31]. Taken together, subthreshold and definite positive psychotic symptoms were experienced by more than 50 % of adolescents and 75 % of adults with 22q11DS. When subthreshold and definite positive psychotic symptoms were examined separately, this revealed that subthreshold manifestations already emerged in few children, whereas definite symptoms were present from adolescence only. In general, hallucinations tended to emerge earlier than delusional ideas (the youngest patient experiencing definite hallucinations was aged 12 years, whereas the youngest patient experiencing definite delusions was aged 15). This goes in favour of Garety’s model of positive symptoms [35], which argues that delusional ideas account for the presence of anomalous experiences and therefore follow the onset of hallucinations in the majority of cases. The description of the phenomenology of positive psychotic symptoms in this population is generally consistent with the previous description of Vorstman et al. [32]. We observed that hallucinations were more frequently auditory than visual and that delusions were predominantly referential. Our results also suggest that combined forms of hallucinations and delusions were almost exclusively present in patients with definite psychotic symptoms and rarely observed in patients with subthreshold manifestations. Altogether, these data suggest that psychosis in 22q11DS should not be considered as an all-or-nothing phenomenon, but rather as a continuum of psychotic experiences of increasing severity. This also underlines the importance of carefully assessing psychotic manifestations in 22q11DS individuals, even in the paediatric population. However, our clinical experience suggests that these patients are often reluctant to share their experiences for fear of being judged negatively, particularly with their families. For this reason, clinicians should always spend time alone with the patient and investigate the presence of symptoms using detailed and oriented questions. In particular, a careful assessment of auditory hallucinations and ideas/delusions of reference should be performed.

The comparison of positive and negative symptoms in participants at different stages of psychosis revealed that the severity of hallucinations, non-persecutory delusional ideas, and grandiosity discriminate between at-risk patients and non-prodromal individuals. Disorganized communication distinguished at-risk patients and individuals with a psychotic syndrome. These results suggest that hallucinations, delusional ideas such as mind reading, ideas of reference, or magical thinking are among the first positive symptoms that are likely to appear in 22q11DS youth. On the contrary, it suggests that more disorganized aspects of the behaviour are more likely to appear during the transition to a full-blown psychotic state. Based on our clinical experience with this population, we recommend instituting an antipsychotic medication to avoid an evolution of symptomatology towards disorganization when psychotic symptoms are present more than once a month or are described as intrusive. By contrast, we observed that the severity of negative symptoms did not discriminate at-risk patients from non-prodromal individuals. There are at least two possible interpretations for this result. On one hand, negative-like symptoms may be a characteristic of the 22q11DS clinical profile and may not be specifically related to the development of psychosis in this population. For example, these manifestations could be related to cognitive deficits. On the other hand, negative symptoms may predict the development of psychosis, while preceding the onset of positive symptoms. This interpretation is commensurate with Dominguez et al. [36], who suggest that negative symptoms appear several years before the onset of positive symptoms and predict their occurrence. Furthermore, negative symptoms have been shown to predict transition to schizophrenia [37, 38]. Future longitudinal studies in 22q11DS examining the predictive value of negative symptoms in the conversion to schizophrenia will help clarify these alternative interpretations.

In accordance with our third hypothesis, we observed that positive psychotic symptoms had different trajectories of evolution over time. Indeed, approximately one-third of the sample experienced emerging symptoms between T1 and T2 evaluations. These patients were below the age of 18 years in 82 % of cases, suggesting that the period of maximum risk for conversion to psychosis is relatively limited in time and occurs during adolescence in the great majority of cases. This should motivate clinicians to perform clinical evaluations at more regular intervals during this period and be particularly vigilant about symptom exacerbation. On the contrary, 20 % of the sample experienced transient psychotic symptoms during the 4-year interval, an interesting finding that has never been reported thus far in this population. As no patient from this subgroup received antipsychotic medication at T1 or T2 evaluations, one cannot assume that symptoms decreased due to a pharmacological intervention. It is either possible that the course of positive psychotic symptoms is naturally transient in some individuals with 22q11DS, or that symptoms were reduced during this period due to an active management of known risk factors for the development of psychotic symptoms (e.g. anxiety management, psychoeducation focused on cannabis use, etc.). Future studies on 22q11DS should further investigate the characteristics of this subgroup of patients, which could give important insights into adequate intervention strategies for the management of psychotic symptoms in this population.

We observed that most cognitive domains were linearly decreasing over time both in patients with and without positive psychotic symptoms. In accordance with other studies [e.g. 39], this suggests that 22q11DS is associated with a general cognitive decline over time, which particularly affects verbal reasoning abilities and processing speed. However, perceptual organization, verbal and visual memory appear to be more stable cognitive domains in this population, given that their trajectory remained constant with age in all participants.

Contrary to our hypothesis, we were not able to identify putative state markers for the onset of psychotic symptoms in this population. Indeed, we did not observe significant difference in the shape of the cognitive trajectories between patients with and without psychotic symptoms. However, and as hypothesized, we observed that several cognitive domains were significantly impaired in patients with psychotic symptoms, but that their developmental trajectory was identical to the one observed in patients without psychotic symptoms (referred to as intercept difference). This suggests that low scores in these domains from childhood are associated with an increased risk to develop psychotic symptoms later in life and may represent early trait markers for psychotic symptoms. In particular, this was the case for global reasoning abilities (full-scale IQ), processing speed, and perceptual organization. Some of these domains have already been reported as risk factors for psychotic symptoms in previous 22q11DS longitudinal studies [13, 14]. It may be the case that children with low cognitive abilities in the context of the 22q11DS [i.e. patients scoring in the range of intellectual disability (IQ <70) vs. patients scoring in the borderline range of functioning or higher (IQ >70)] have more difficulties coping with their environment and experience it as being more unpredictable [40]. This, in turn, could predispose to the development of anxiety, which is a known important risk factor for schizophrenia [4]. However, this interpretation remains speculative and should be tested in future studies.