The short report by Gunawardana et al ^{Reference Gunawardana, Davey Smith, Zammit, Whitley, Gunnell and Lewis1} succinctly argues that a short inter-pregnancy interval, a proxy for fetal undernutrition and stress, increases the offspring's risk of later schizophrenia. The authors hint at a causal relationship. This is compelling because it suggests that an affordable public health intervention via the promotion of dietary supplements in the postpartum period may later reduce schizophrenia prevalence.

Although the authors compare pre- and post-birth intervals and adjust for a number of confounders, their findings may still relate to bias and residual confounding. First, the timing of schizophrenia measurement may distort the prevalence and gender ratio of schizophrenia. This is important because a short inter-pregnancy interval is known to favour male offspring. ^{Reference Greenberg and White2} Looking at the cohort's median year of birth (1978) and the latest possible date of outcome measurement (2002), an individual's lifetime history of schizophrenia would be recorded at 24 years. As there is a significant gender variation in the age-specific incidence of schizophrenia, ^{Reference Thorup, Waltoft, Pedersen, Mortensen and Nordentoft3} the median cohort age of 24 years is likely to bias the cohort towards male schizophrenia prevalence and overestimate the predictive validity of the short inter-pregnancy interval.

Second, the finding of no relationship between the post-birth inter-pregnancy interval and later schizophrenia does not discount residual confounders, including ethnicity and genetic factors, from contributing to the study's main findings. Genetic and familial factors, including ethnicity, are both associated with short inter-pregancy intervals ^{Reference Rodgers, Kohler, Christensen, Rodgers and Kohler4,Reference Rawlings, Rawlings and Read5} and schizophrenia. ^{Reference Fearon, Kirkbride, Morgan, Dazzan, Morgan and Lloyd6} The current study did not mention adjusting for offspring ethnicity, although its design would make it possible. However, any epidemiological study would struggle to separate the prenatal effect of the inter-pregnancy interval from maternal–child genome sharing.

Epidemiological designs will only drive hypotheses so far in examining the causal relationship between prenatal micronutrient depletion and later psychopathology. That said, there would be scientific value in examining cohorts pre- and post-introduction of public health recommendations of periconceptional folic acid vitamin supplementation. In addition, further work analysing the correlates of prenatal nutrient depletion as additive risk factors could provide further evidence of a dose–response relationship. For example, are the risks of schizophrenia enhanced when there is a history of short pre-birth interval plus prior multiple births, concurrent breastfeeding or postnatal vitaminosis?

Introducing postnatal vitamin supplementation to reduce schizophrenia prevalence is an enticing idea; however, it would be important to use a variety of research designs to establish or exclude causality before implementing any change in public health policy.