Schizophrenia is a complex, chronic, and disabling illness that presents with heterogeneity in its clinical appearance, in patterns of psychopharmacological response and in long-term outcomes [1, 2]. While the last few decades of research have given rise to a tremendous wave of interest regarding the natural illness course of schizophrenia, the overarching goal of this research of influencing the prognosis and outcome for schizophrenia patients remains pertinent today. The validity of possible predictors of treatment response and long-term outcome in schizophrenia patients has been a topic of much study in recent years. However, the literature still lacks a clear-cut picture regarding which factors are valuably prognostic in the clinical management of schizophrenia spectrum disorders.

Numerous empirical studies concentrate on the predictive values of the variables detectable at the first episode of schizophrenic illness for the long-term patient outcome. Some practical factors, including sex and age at onset, have been reported as being among the most important determinants of the outcome of schizophrenia [3]. A number of studies have suggested that women in general may have a later age of onset and an overall lower severity of illness, suggesting a protective effect of estrogens [4]. Several negative predictors of outcome in schizophrenic patients have been identified, including male sex [5], early and non-acute onset of syndromes [6], the prevalence of negative symptoms [7, 8], the presence of affective symptoms [9], poor pre-morbid functioning [10], and a delay in starting pharmacological treatment [11]. All of these predictors have been associated either with poorer long-term global functioning or with higher rates of relapse or hospitalisation. In particular, patients with schizophrenia with earlier ages of onset are more likely to be males and to have poor pre-morbid adjustment, lower educational achievement, more evidence of structural brain abnormalities, more prominent negative symptoms, more cognitive impairment, and a worse overall outcome [12] as well as a higher likelihood of having relatives with schizophrenia [13, 14].

The onset of schizophrenia prior to age 13 is exceedingly rare [15], but an estimated 39% of males and 23% of females with schizophrenia develop the illness by the age of 19 [16]. Patients with early-onset schizophrenia (EOS; onset by age 18) [17], including childhood-onset schizophrenia (COS; onset by age 13) [16] and adolescent-onset schizophrenia (AOS; onset after age 13 and before age 18) show a number of the same neurobiological abnormalities observed in adult-onset schizophrenia, suggesting the involvement of a common neurobiological substrate [17, 18]. However, compared to patients with adult-onset schizophrenia, an early onset of schizophrenia appears to be associated with higher rates of pre-morbid abnormalities [17, 19], worse cognitive performance [20] and worse functional outcomes [21]. Taken together, these studies indicate that EOS may result in a more severe form of the disorder.

While age of onset and sex have been documented to be fundamental for understanding schizophrenia and bipolar disorder separately [3, 22], there is also substantial clinical and neuropsychiatric overlap between these two disorders. We have embodied this overlap in schizoaffective disorder, a diagnostic category with features of both schizophrenia and affective disorders but with poor construct validity [23]. The current diagnostic system has been further called into question by recent studies indicating shared genetic determinants of bipolar disorder and schizophrenia [24]. In spite of these overlaps, there has been little research examining the characteristics and symptoms of psychosis within both schizophrenia and bipolar disorder, which could yield evidence of commonalities as well as differences among patients with psychosis [25].

In this study, we estimated the relationship of age of onset to other clinical characteristics in well-classified patients diagnosed with schizophrenia spectrum disorders. We also compared psychopathology and other clinical variables between schizophrenia and schizoaffective patients with early and adult onset. With regard to the age of onset and classification of types of schizophrenia or schizoaffective disorder, previous studies have adopted cut-off points. For instance, a review of adolescent research used the age of 17 to distinguish child- and adolescent-onset from adult-onset schizophrenia [26]. Because 17 years of age is the cut-off found in most studies and clinical settings [26], we selected 17 years as the cut-off for early-onset schizophrenia.

Among the numerous clinical characteristics used to clarify the schizophrenia spectrum disorders, the age of illness onset is widely accepted as having particularly powerful clinical and prognostic significance. The complexity and variety of effects of the age of onset in schizophrenia patients reported in the literature are due not only to the difficulty in operationally defining the age of illness onset but also to the broad distribution of ages of onset from preadolescence to later adulthood. In this cross-sectional study, there was evidence for statistically significant relationships between age of onset and cognitive impairments and impulsivity traits in this group of schizophrenia spectrum disorders. Patients with early onset had higher levels of cognitive impairments and impulsivity traits than did patients with adult onset. This is compatible with the generally accepted view of early-onset cases as having unique clinical and prognostic consequences. However, we do not have any evidence for a causal relationship between age of onset and cognitive impairments and impulsivity traits. No definitive conclusion can be drawn until longitudinal prospective studies are carried out.

The mean age of onset for all schizophrenia patients who participated in this study was slightly older than is generally reported in populations of schizophrenia patients [38], particularly when recorded as the year of life when the subject first met DSM-IV [12] criteria. A possible explanation is that a large proportion of our patients (about 70%) presented with the paranoid type of schizophrenia, which is characterised by a considerably older age of onset (mean age of 28.5 years vs. 19.9 years in patients with non-paranoid schizophrenia). The present results also show that schizophrenia patients were not different from patients with schizoaffective disorder in terms of age of onset. There is a growing body of research specifically regarding early-onset schizophrenia [16–18], but research regarding youths with schizoaffective disorder is sparse [39]. In fact, most studies include schizoaffective disorder as an exclusionary criterion or combine both diagnoses into one group for data analysis [39]. Further complicating matters is the fact that these diagnoses are often contingent on a longitudinal illness course, yet diagnosis is generally made using cross-sectional information. The DSM-IV diagnostic criteria for schizoaffective disorder require that mood episodes be present for a substantial portion of the duration of the illness [12]. This diagnostic assignment may change over time as the course and presentation of psychotic symptoms become obvious [39, 40]. For example, in the clinical setting, a patient's diagnosis can change from schizophrenia at baseline to schizoaffective disorder at discharge [40]. Further research will be needed to help clinicians distinguish schizophrenia from schizoaffective disorder in patients with early-onset schizophrenia. Despite this, one of the most salient and overarching findings of our study is that the patients with schizophrenia and schizoaffective disorder are more similar than different in terms of their demographic and symptom profiles. Our findings provide additional support for shared etiological and pathophysiological features across schizophrenic disorder groups. Such information will have important prognostic and treatment implications.

Researchers have shown that age of onset may not necessarily act as a unique determinant in the course of schizophrenic disorder as evidence indicates that men have an earlier age of illness onset than women [3–6] and a more severe course of illness, particularly in the short and medium terms [41]. A remarkable finding in the present study was that we were not able to establish differences between male and female patients in demographic variables, including age at onset and symptom severity, or in total scale or subscale scores. This is in contrast with previous studies that found symptomatic differences between the sexes. In these previous studies, negative symptoms were consistently found to be more severe in men [42]. This discrepancy might be due not only to differences in the rating scales applied but also to sample differences. The lack of evidence for a sex difference is difficult to explain. Our patients' mean age (mean age = 40.57 years) was higher than those of other studies [43], the female patients (mean age = 41 years) were older than the males (mean age = 40 years), and female patients (mean duration = 15.25 years) had a longer illness duration than males (mean duration = 14.94 years); this might be related to a putative progressive reduction in symptom differences. The results indicate that the differences in clinical characteristics of schizophrenic disorders between early- and adult-onset patients may be more pronounced than those between patients of different sexes, but an interaction effect might be present between sex and age of onset.

A number of independent cognitive deficits were apparent in our chronic schizophrenic patients, especially in early-onset cases. The findings of the current study and the study of Hoff et al. (1992) indicate a more generalised, diffuse cognitive deficit in chronic schizophrenic disorders [44]. Our results also support the assertion of the DSM-IV (1994) that schizophrenia patients with younger ages of onset are more cognitively impaired [12]. It seems that an earlier onset of schizophrenia is associated with a more severe course irrespective of duration of illness [45]. Given its cross-sectional nature in this study, however, no conclusion can be drawn regarding causality and alterative explanations of the findings cannot be ruled out. For instance, it is possible that patients with adult onset had better response to antipsychotic medications, thereby reducing their severity of symptoms. Specifically, we used the cognitive component of the PANSS to assess cognitive function in patients with schizophrenia. It has been documented that higher scores on the PANSS cognitive component are significantly correlated with poorer performance on neuropsychological tests [46].

Action without planning or reflection is central to most definitions of impulsivity. In the present study, we used the BIS questionnaire, which tends to measure impulsivity as a stable characteristic, as a self-reported assessment of impulsivity [30]. There were significant associations between early age of onset and severity of impulsivity traits. Previous reports have suggested that schizophrenic patients are likely to exhibit impairments on a wide range of neuropsychological tasks, including attention and executive functioning [47]. Heaton et al. (2001) showed that the neuropsychological impairment in patients with schizophrenia appeared to remain stable regardless of baseline characteristics and changes in clinical state [48]. Reduced P300 amplitude, a neurophysiological parameter associated with impulsivity and behaviour disinhibition [49], and a P300 effect size (d) that was smaller in amplitude and longer in latency have been observed in schizophrenic patients compared to normal controls, with the strongest effects obtained from the auditory oddball task [50]. Therefore, it is plausible that psychopathological and neurocognitive impairments in schizophrenic patients are mediator variables responsible for the effect of impulsivity on the age of onset in schizophrenia spectrum disorders. The results reported in the present study support this relationship. Because impulsivity is present as a relatively stable trait, it would appear that greater impulsivity is already present at onset in early-onset schizophrenia. However, no definitive conclusion can be drawn until further prospective studies are performed.

Compared with the regression model, significant effects of age at onset were found in the negative symptom component, cognitive component, and the total score, but not in the positive component of the PANSS in the ANCOVAs. Patients with early illness onset scored higher on negative symptoms, cognitive symptoms, and general psychopathology than did patients with adult-onset illness. To further evaluate the magnitude of the predicted difference, an effect size test was conducted. The standardised effect size difference for cognitive impairment between groups was 0.387, reflecting a medium-sized effect [51]. Moreover, the standard effect sizes for negative symptoms and impulsivity traits were 0.427 and 0.511, respectively, also reflecting medium-sized effects [51]. However, the standardised effect size difference for positive symptoms between groups was 0.121, reflecting a smaller effect size [51]. These results agree with those of some previous systemic studies [52]. Similarly, some studies have reported that negative thought disorder was less severe in patients with older ages of illness onset. However, there was no effect of age of onset on the depressive symptoms in this study, a finding that is consistent with other comprehensive studies [4, 52]. Together, given the exploratory nature of these studies, these data suggest that any phenomenon related to age of onset in schizophrenia based on these preliminary findings should be treated with caution.

In the present study, considering the results of t-tests for differences between the early-onset and adult-onset groups, it was expected that education would be related to age of onset as patients should have completed less schooling if their first episode occurred when they were still attending school. A likely explanation is that our patients with earlier ages of onset had poor educations due to the cognitive dysfunction associated with poorer outcomes in early-onset schizophrenia [12, 53]. However, this significant effect was appreciably reduced after controlling for illness duration and current age. The findings suggest that the difference in the educational levels between the two groups may be strongly affected by illness duration and current age. It is difficult to estimate the confounding influence of illness duration on our test results due to the retrospective design.

It is uncertain, however, whether the effects of age of onset found in the present study reflect qualitatively specific schizophrenia or merely quantitative differences in psychopathology and impulsivity between early- and adult-onset illness in our patients. One recent study reported that the relationship between an older age of onset and less severe negative symptoms is also present in chronically ill schizophrenic patients with an age of onset of younger than 45 years [54]. Thus, future research is needed in this area, particularly concerning the potential consequences of age of onset, utilising different clinical measures (especially as the results indicate that early onset is a risk factor) and a broader array of measures tor precisely define the course of schizophrenic disorders.

In stabilised schizophrenic patients, assessment of subjective QOL has good reliability and concurrent validity [55, 56]. Hence, measurement of subjective QOL may be considered as a pertinent indictor of the state of health of stabilised schizophrenic patients [56]. The present study tested the relationship of age of onset with the QOL of patients with schizophrenia spectrum disorders by using t-tests. The results showed that patients with an early onset of schizophrenia spectrum disorders were likely to have worse QOLs than those with an adult illness onset. A partial explanation for this may lie in the fact that an early onset of illness has been found to be a predictor of unfavourable prognosis and is correlated with higher global severity [57], higher rates of chronicity, and more probable impairments in cognitive performance [58]. However, this significant effect was largely reduced after controlling for illness duration. The findings suggest that the difference in QOL levels between the two groups may be strongly affected by illness duration. Besides, in patients with schizophrenia, adaptation and significant improvement in subjective QOL may be assumed to occur at a later stage of illness [59]. This finding is compatible with the results of our study, which showed that older patients are more satisfied with their lives than younger patients are. (Pearson's r = 0.218, p < 0.01).

There are some limitations of our research. First, only inpatients in the chronic setting were recruited in the present study. The results could not demonstrate whether the effect of age of onset as measured in our study indicates a trait or state characteristic. Additionally, we could not generalise our findings to all schizophrenia subjects. Thus, replication of the current findings in stabilised outpatients will be necessary. Second, because the present study required informed consent and included psychopathological assessments, we did not include subjects who were very uncooperative. Thus, we lack demographic characteristics of non-volunteers. However, it should be noted that those uncooperative subjects were demographically different from the volunteers, and thus the influence of our results might be limited. Third, as noted above, the size of the early-onset group was relatively small, which likely limited our ability to detect group differences due to low statistical power, but this may reflect a greater prevalence of adult-onset schizophrenic disorder cases [16]. Fourth, it is important to emphasise that methodological problems such as the retrospective design limit our interpretation. All data on the illness course, however, were based on information documented at the time of inpatient treatment, including the age of onset of the first psychotic episode and other demographic and clinical characteristics, which can be biased by recall effects. Thus, a prospective comparison of characteristics at illness onset of patients with schizophrenia spectrum disorders will be necessary for future research. Finally, given the retrospective design of our study, the psychopharmacological variables were not controlled a priori, and thus it was not possible to determine the effects of the medications on some aspects of cognition and the clinical course of the disease.

In summary, the present study showed that the variance of demographic and clinical characteristics in schizophrenia spectrum disorders may be greater between early- and adult-onset patients than between patients of different sexes. The findings are roughly in line with Howard's (2000) results [60]. Our statistical analyses also demonstrated that these significant associations were not accounted for by the duration of illness. Thus, the ability to recognise psychopathological symptoms of schizophrenia to start psychopharmacological and psychosocial interventions as soon as possible is becoming a central issue in clinical practice [61]. Accordingly, further investigation is needed to gain a better understanding of the age effect on the illness process of schizophrenia spectrum disorders.