Evidence of a Cohort Effect for Age at Onset of Schizophrenia

The term “cohort effect” is used to describe variations in the characteristics of an illness (such as the incidence or the age at onset) over time among individuals who are defined by some shared temporal experience, such as year of birth.

Such a cohort effect has been described for affective and schizoaffective disorders (1, 2). Community and family studies have reported an increase in the prevalence of these disorders and a decrease in age at onset in the cohorts born after World War II.

In schizophrenia, the influence of secular trends has been described in the incidence of the disease (3). However, to our knowledge, the existence of a cohort effect for age at onset of schizophrenia has never been studied in a community sample. If such a cohort effect for age at onset of schizophrenia exists, it may have contributed, as a bias, to the decrease in age at onset of schizophrenia reported within succeeding generations of families, also called the anticipation phenomenon (4).

The present report analyzes three successive birth cohorts of schizophrenic subjects to determine if age at onset has changed over time.

Method

Data were obtained from medical records of all patients admitted to a psychiatric hospital during the years 1976, 1986, and 1996 for schizophrenia or other psychotic disorders (delusional or schizoaffective disorders).

Medical records of 877 patients, born between 1905 and 1984 in Normandy, were analyzed. Data for each individual included information on gender, date of birth, and date of first admission. Diagnoses of schizophrenia (per DSM-IV criteria) and determination of age at onset were made retrospectively through medical record review by the same psychiatrist (C.D.M.), who was blind to patient identity. Two onset points were studied: age of first appearance of psychotic symptoms and age at first hospitalization (5).

Information about alcohol or substance abuse at age at first appearance of psychotic symptoms as well as psychiatric morbidity in the relatives of the patients were obtained from medical records. Schizophrenic patients who had at least one first-degree relative affected with a psychotic disorder (schizophrenia or schizoaffective or delusional disorder) were classified as having a family history. A patient who reported that he or she did not use substances was considered as being a nonuser.

From the 877 medical records examined, only subjects with a confirmed DSM-IV diagnosis of schizophrenia and age-at-onset data were included in these analyses (N=419). Patients born before 1945 were gathered together in order to balance the study group sizes. The 419 patients were placed into one of three birth cohorts: 1905–1944 (N=96), 1945–1964 (N=200), and 1965–1984 (N=123). The remaining subjects were excluded either because they did not meet DSM-IV criteria for schizophrenia (N=421; 150 born in 1905–1944, 188 born in 1945–1964, and 83 born in 1965–1984) or determination of age at onset was not possible because of insufficient data in medical charts (N=37; 15 born in 1905–1944, 17 born in 1945–1964, and five born in 1965–1984).

For statistical analysis, we first used a two-way analysis of variance to compare age at onset by gender and by birth cohort. In a second step, we used a nonparametric test (Kruskal-Wallis chi-square test) followed by a Bonferroni post hoc test to compare median age at first appearance of psychotic symptoms and median age at first hospitalization (BMDP, Statistical Software, Berkeley, Calif., 1990). A chi-square test was used for comparison of qualitative variables.

Results

The number of schizophrenic patients with a comorbid substance abuse diagnosis increased significantly across the three successive birth cohorts, as did the male-female ratio (Table 1). The mean age at first appearance of psychotic symptoms and age at first hospitalization significantly decreased in successive birth cohorts. There was no gender effect (age at first appearance of psychotic symptoms: F=0.05, df=1, 413, n.s.; age at first hospitalization: F=0.83, df=1, 413, n.s.) or birth cohort-by-gender interaction (age at first appearance of psychotic symptoms: F=0.07, df=2, 413, n.s.; age at first hospitalization: F=1.18, df=2, 413, n.s.). When median ages for the successive birth cohorts were considered for both first appearance of psychotic symptoms (24.5, 22, and 20 years, respectively) and first hospitalization (26, 24, and 21 years), a significant decrease was observed for both (age at first appearance of psychotic symptoms: χ²=48.0, df=2, p<0.0001; age at first hospitalization: χ²=44.5, df=2, p<0.0001).

The time elapsed between age at first appearance of psychotic symptoms and age at first hospitalization decreased slightly but not significantly between the three successive cohorts (mean=1.76 [SD=4.7], 1.71 [SD=3.6], and 1.37 [SD=2.3] years, respectively) (F=0.04, df=2, 413, n.s.); there was no gender effect (F=1.22, df=1, 413, n.s.).

Discussion

We observed a progressive decline of about 5 years in both age at first appearance of psychotic symptoms and age at first hospitalization in three successive birth cohorts of schizophrenic patients (spanning 1905 to 1984). To our knowledge, this is the first report of a significant birth cohort effect for age at onset of schizophrenia.

The results of this study should be, of course, interpreted in the light of the methodological limitations inherent to retrospective investigations that use hospital admission rates. Kendell et al. (6) considered that some schizophrenic patients may not be referred to psychiatrists or may never be admitted to the hospital. This is less likely to occur in France, since psychotic patients are usually referred to psychiatric hospitals, which have offered the most suited care since the turn of the century.

In our study, when men and women were examined separately, the cohort effect observed for age at onset remained unchanged in either group. By contrast, Kendell et al. (6) observed that mean age at first hospitalization for schizophrenia decreased by 4 years between the 1970s and 1980s, especially in women.

Corresponding to the decrease in age at first appearance of psychotic symptoms or age at first hospitalization, cohort trends have been described for a number of important factors possibly related to schizophrenia, including the use of alcohol and illicit drugs. As observed in our study, O’Malley et al. (7) reported a higher rate of drug abuse in younger cohorts of schizophrenic patients. When schizophrenic patients who fulfilled criteria for substance abuse were excluded from the study, the cohort effect observed remained. However, information about substance use came from chart review, was not corroborated, and must be viewed cautiously.

In multiply affected families that exhibit clinical anticipation (8, 9), such a cohort effect may have contributed to the decrease in age at onset observed in successive generations. In order to study this potential bias, Heiden et al. (10) found, in two nonoverlapping small birth cohorts of schizophrenic patients admitted to the hospital, that the mean age at first hospitalization was slightly but not significantly higher in the older cohort. In our study, the exclusion of patients with a family history did not change the cohort effect observed.

Ongoing research regarding family aggregation of schizophrenia needs to incorporate cohort effects into genetic models.

TABLE 1

Enlarge table