This issue includes several papers that examine relationships between developmental processes and psychiatric illness. All of these studies follow subjects across time. A paper by Kraemer and co-workers reviews how misleading cross-sectional designs can be and how advantageous longitudinal designs can be in psychiatric research. Using examples primarily from studies of Alzheimer’s disease, they illustrate pitfalls in cross-sectional research that are comprehensible to clinicians. Recent experience in a prospective magnetic resonance imaging study of brain development in childhood and adolescence has underscored the power of longitudinal research designs. In one study, nonlinearity of regional cortical development was evident only from longitudinal data, despite the presence of cross-sectional data for a much larger sample (1). Here the issue was the enormous true individual variance of these brain measures, which decreased the power of cross-sectional analyses.

The remaining reports on developmental processes in this issue highlight two quite different themes. The first is the risks and benefits of drug treatment during pregnancy and the postpartum period. Viguera et al. show that lithium discontinuation during pregnancy is associated with a relapse rate similar to that for lithium discontinuation among nonpregnant patients with bipolar disorder followed over a similar period of time (52% and 58%, respectively). However, the postpartum recurrence is much higher (80%), even for patients who remain stable after lithium discontinuation throughout pregnancy. Given lithium’s well-documented fetal toxicity, a trial discontinuation period at least early in pregnancy would seem generally indicated, as would the need for resumption of the medication postpartum. In many instances of major psychiatric illness, nursing mothers will be maintained on their psychotropic drug regimen. Stowe et al. report on a systematic study in which levels of paroxetine in breast milk and infant serum were examined in relation to maternal paroxetine dose. Encouragingly, infant serum paroxetine was not detectable, and no adverse effects were seen. This unusually well-done study extends a series of earlier reports supporting the safety of selective serotonin reuptake inhibitors during lactation.

A second theme is the association between pre/perinatal adverse events or stressors and adult psychiatric outcomes. Brown and coworkers extend their previous study of fetal exposure to famine and major affective disorder in adulthood. Newly ascertained cases from their cohort of pregnancies during the Dutch Hunger Winter of 1944–1945 show an association between exposure to famine during the second and third trimesters and adult affective disorders. These effects held for both unipolar and bipolar disorders. The findings extend and support a growing body of data from a British cohort (2) and other studies documenting a variety of early neurodevelopmental precursors to adult affective illness. Both direct effects (nutritional impairment of late fetal brain development) and indirect effects (mediation through maternal stress) are plausible. What are we to make of the current findings of Brown et al., which are similar to their previous findings of famine and later schizophrenia in the same cohort? And what are we to make of the growing body of other studies discussed by these authors that have found similar subtle neurodevelopmental precursors to affective illness? Are the causative factors truly nonspecific or are our measures too gross to differentiate precursors for these quite different conditions? The probable nonspecificity of pregnancy or perinatal risks is underscored by recent pediatric studies. A variety of other seemingly very different adult medical disorders, including adult coronary heart disease and stroke (3, 4) and hyperinsulinemia in adult years (5), all with genetic components, have been associated with smaller growth in utero. Conversely, women with undiagnosed genetic thrombophilia have a greatly increased rate of complications of pregnancy (6), but the level of maternal screening needed to detect this condition is rarely if ever carried out in birth cohort studies.

Two papers report associations of fetal/neonatal complications with the risk of schizophrenia 19 years later (Zornberg et al.) or with differences in brain structure sizes in monozygotic twin pairs discordant for schizophrenia (McNeil et al.). Zornberg et al. used systematic prospective pre- and perinatal data from the National Collaborative Perinatal Project to derive a measure believed to represent hypoxic-ischemic mediating events. The researchers found a highly significant doubling of the risk of adult-onset psychosis in the presence of these risk factors, compared with no relevant complications, a relationship that was particularly strong for schizophrenia and other nonaffective psychoses. The authors are appropriately cautious in their interpretation because of the complexity of the risk factors they identified.

The association between chromosomal or genetic abnormalities and obstetric hazards complicates the interpretation of data that demonstrate relationships between obstetric hazard and later psychosocial adjustment, because a causal relationship cannot be specified without additional information (7). Even outcomes as seemingly clear as cerebral palsy, which was for a long time believed to be caused by pregnancy and birth complications, are now believed to be secondary to disturbances very early in fetal development (8, 9). Similarly, maternal genetic thrombophilia has a significant association with preeclampsia (6).

The report of McNeil et al. on the relationship of obstetric complications and brain structure size differences in monozygotic twin pairs discordant for schizophrenia provides yet another demonstration of the power of twin pair research for psychiatric genetics. As anticipated, the ill twin consistently had relatively smaller hippocampi and larger brain ventricles. More intriguingly, these abnormalities in the ill twin were significantly related to labor-delivery complications, particularly prolonged labor. These findings suggest not only that there may be different timing sequences for obstetrical complications but also that different etiological influences may be operating within pairs. It is puzzling, however, that long labor emerged as a risk factor in these analyses, as it has not been an important risk factor for other neurologic outcomes (8, 10)

The information age, with its explosion of technology for storing and merging data sets, together with the power of longitudinal research designs facilitates developmental studies. Developmental hypotheses, whether they address schizophrenia, affective disorder, or nonpsychiatric disorders, will benefit many areas of medicine. The wealth of positive findings calls attention to possible mechanisms but does not truly explain any one illness. The nonspecificity of these findings may ultimately be sorted out by using improved risk measures or by developing better diagnostic classifications. On the other hand, the very generality of these risk measures may give them more public health usefulness, as information they provide may help prevent several outcomes.

Dr. Rapoport is Chief, Child Psychiatry Branch, National Institute of Mental Health. Address reprint requests to her at Bldg. 10, Rm. 6N-240, 9000 Rockville Pike, Bethesda, MD 20892.