Dr. Verdoux and Colleagues Reply

To the Editor: The statistical model we used was intended to explore the clinical correlates (i.e., age at onset, gender, family history) of a history of obstetric complications in schizophrenia. The model used by Schaub and colleagues addressed another question and was designed to examine the explanatory variables of age at onset in schizophrenia. The main limitation of this latter model is that it does not allow the assessment of a dose-response relationship between age at onset and a history of obstetric complications. Our model showed that the probability of having had a history of obstetric complications is diminished with an increase of age at onset. This finding was reported more simply in our conclusion as indicating that subjects with a history of obstetric complications are more likely to present with early onset schizophrenia than those without such a history. Whatever the formulation, the interpretation of this finding is similar, i.e., it demonstrates an association between age at onset and obstetric complications. We are puzzled by the second point raised by Schaub and colleagues—that information was lost in our statistical analyses—since we used precisely the method that is described in their letter. We specified that regression methods were used to obtain pooled weighted estimates of the odds ratios from the individual patient data, and a variable, “study,” was encoded to indicate from which study the data were obtained.

We were most interested in the findings reported by Schaub et al., which demonstrate that both family history and obstetric complications are independently associated with age at onset in schizophrenia. These results are in accordance with those of previous studies examining the relationship between family history and age at onset (1); they also confirm the association between early onset and obstetric complications. We have now examined the relationship between age at onset and the number of definite obstetric complications (categorized into a four-level variable) in the pooled sample of subjects with schizophrenia. The proportion of subjects with onset before 21 years was higher in subjects with three and more obstetric complications (68.8% of 16) than in subjects with two obstetric complications (64.0% of 25), one obstetric complication (54.5% of 101), and no obstetric complications at all (46.7% of 323). A significant linear trend was found in the association between the dependent variable, age at onset (categorized into a binary variable, <22 versus 22 years), and explanatory variable, the four-level variable “number of obstetric complications,” in a model fitted with the variable “study” (weighted average odds ratio for linear trend=1.32; 95% confidence interval=1.0–1.69; likelihood ratio=4.03, df=1, p=0.04). There was no evidence for heterogeneity across different studies. This result indicates that the higher the number of definite obstetric complications, the more likely the onset before 21 years. This dose-response relationship adds further evidence supporting the hypothesis that obstetric complications are involved in the pathophysiology of early-onset schizophrenia.