To the Editor: Hélène Verdoux, M.D., and her colleagues R7315511CHDCCHHE show a linear trend for the association of age of onset in schizophrenia and a history of obstetric complications.

In general, we agree with their conclusions and provide data supporting their results, but we were concerned with their modeling strategy. Taken seriously, it is shown that earlier age at onset of schizophrenia increases the probability of having had obstetric complications, while the real conclusion drawn is that obstetric complications decrease the probability of earlier age at onset.

To put the model in the correct order, one had to change response and explanatory variable. The difference in estimated parameters may be small, albeit they are not identical, depending on the covariance terms in the model matrix.

Further, the use of statistical techniques for meta-analysis of separate studies might result in loss of information for pooled data of individual patients. Since these data were available, one should consider using the pooled data only and including a variable indicating from which study the data came.

For our own analysis, we extracted records from the AMDP database R7315511CHDEBAJI of the Department of Psychiatry, Free University of Berlin, where patient description with the AMDP system is part of mandatory clinical routine. Data from 827 schizophrenic patients (ICD-9 codes: 295.0–295.6) treated between 1981 and 1995 were eligible, i.e., information concerning age at onset of schizophrenia, obstetric complications, and affected family members was available. The sample consists of essentially unselected hospital patients. To make results comparable, we dichotomized age at onset as either 18 or 21 years, younger versus older.

Covariates were sex, a familial history of schizophrenia, and the presence of any obstetric complication. Using logistic regression, we obtained odds ratios of 1.76 (95% confidence interval: 1.03–3.02; p=0.04) for obstetric complications and of 0.61 (95% confidence interval=0.41–0.90; p=0.01) in favor of the female sex, while for familial history no significant result was found. Changing the age-at-onset criterion to 21 years and younger yielded an additional effect for familial history (odds ratio=1.68; 95% confidence interval=1.07–2.63; p=0.02) of about the same magnitude as for obstetric complications (odds ratio=1.61; 95% confidence interval=1.02–2.54; p=0.04), the effect of sex remaining about equal (odds ratio=0.54; 95% confidence interval=0.40–0.73; p<0.001). No significant interaction terms were found.

Data originating not from planned studies, but from clinical routine will suffer from some unreliability due to unstandardized measurement. Nevertheless, we think that even if mild complications were missed, at least moderate and severe types of obstetric complications were recognized, thus explaining the somewhat higher odds ratios in our study group. Consistent with several studies, a familial history of schizophrenia also was associated with an earlier age at onset (under 22 years), although not with an age at onset under 19 years.