Subjects were followed up for a mean of nine years after the age of 16 years. During this follow up period 639 (0.09%) were admitted with a diagnosis of schizophrenia and 1311 (0.18%) with a diagnosis of non-affective non-schizophrenic psychosis. The estimated annual incidence rates were 0.10 and 0.21 per 1000 person years, respectively.
shows the characteristics of subjects in relation to the age of their father. People with older fathers tended also to have older mothers and mothers who had had more pregnancies. People with older fathers were more likely to lose their parents before they reached the age of 18 years; they were more likely to have a family history of psychosis; and their parents were more likely to come from extreme ends of the distribution of the various measures of socioeconomic position.
Characteristics of subjects according to paternal age groups. Figures are number (percentage) unless stated otherwise
shows the hazard ratios for successive five year paternal age groups. In the basic model, adjusted for sex, age, and maternal age, there is a clear increase in the risk of schizophrenia with increasing paternal age. There is a suggestion that the offspring of younger fathers (< 21 years) are also at somewhat greater risk than those with fathers aged 21-24 years (hazard ratio 1.34, 95% confidence interval 0.74 to 2.43), but there was no evidence of a non-linear association between paternal age and schizophrenia (P = 0.39, quadratic). The strength of the association between paternal age and schizophrenia changed hardly at all when we also controlled for risk factors related to birth, parental death before the subject reached the age of 18 years, and family history of schizophrenia, though there was a modest reduction in models after we controlled for the three socioeconomic variables. The variable mostly responsible for this drop was highest annual parental income.
Hazard ratios (with 95% confidence intervals) of schizophrenia in relation to paternal age (n=639 cases of schizophrenia)
shows the results for the same series of models for non-schizophrenic non-affective psychosis. Overall, we found a much weaker association. Just as with schizophrenia we found an increased risk in the offspring of younger men (< 21 years) with some evidence for a non-linear association (P = 0.01, quadratic).
Hazard ratios (with 95% confidence intervals) of non-schizophrenic non-affective psychosis in relation to paternal age (n=1311 cases of non-schizophrenic, non-affective psychosis)
When we combined cases of narrowly defined schizophrenia and non-schizophrenic non-affective psychosis into one category the hazard ratio associated with each 10 year increase in paternal age was 1.31 (1.17 to 1.47) when we controlled for sex, age, and maternal age and 1.23 (1.10 to 1.37) in the fully controlled model.
Maternal age showed a weak association with schizophrenia in the unadjusted model (hazard ratio per 10 year increase in maternal age 1.21, 1.02 to 1.43), but this association disappeared as soon as we controlled for paternal age, sex, and age (0.83, 0.64 to 1.07) (see table B on bmj.com
). We found a similar pattern for non-schizophrenic non-affective psychosis (crude hazard ratio per 10 year increase in maternal age was 1.10 (0.98 to 1.23) and 0.94 (0.79 to 1.12) when we controlled for sex, age, and paternal age
We calculated that the population attributable fraction of schizophrenia in this sample due to having a father aged > 30 years at birth was 15.5%.
Association with family history, sex, birth weight, Apgar scores, and age of onset
Associations with paternal age differed in those with and without a family history of schizophrenia (P = 0.04 for interaction). The hazard ratio for every 10 years of paternal age was 1.60 (1.32 to 1.92) in those with no family history of schizophrenia and 0.91 (0.44 to 1.89) in those with an affected first degree relative.
We found no strong evidence for any marked difference in the association between paternal age and schizophrenia by sex (P = 0.20, for interaction). Hazard ratios increased by 1.65 (1.32 to 2.06) for men and by 1.40 (1.03 to 1.89) for women for every 10 years of paternal age.
We found no evidence for any marked difference in associations in those with high and low birth weight (P = 0.15 for interaction). The association of paternal age with schizophrenia, however, differed in relation to the subjects' Apgar score at birth (P = 0.02 for interaction). In those with normal Apgar scores (7-10) at one minute hazard ratios increased by 1.60 for every 10 years of paternal age (1.33 to 1.92) but not in those with Apgar scores suggestive of the need for resuscitation (hazard ratios decreased by 0.92, 0.31 to 2.76). Similar but weaker effects (P = 0.16 for interaction) were seen with Apgar scores at five minutes. We found no marked change in the strength of association between paternal age and schizophrenia when we restricted the proportional hazards model to the first or the second half of the follow up period.