Women with schizophrenia frequently discontinue antipsychotic medications during pregnancy. However, evidence on the risk of postpartum relapse associated with antipsychotic use during pregnancy is lacking.

To investigate the within-individual association between antipsychotic continuation during pregnancy and postpartum relapse in women with schizophrenia.

This retrospective cohort study used data of women with schizophrenia who gave live birth between 2007 and 2018 identified from the National Health Information Database of South Korea. Women were classified according to antipsychotic use patterns during the 12 months before delivery as non-users, discontinuers and continuers. Relapse was defined as admission for psychosis (ICD-10, F20–29). The incidence rate ratio (IRR) for admission for psychosis in the 6-month postpartum period was estimated using conditional Poisson regression, with the reference period set between 2 and 1 years before delivery. Additionally, we calculated the relative risk ratios (RRRs) for the IRRs of different antipsychotic use patterns.

Among the 3026 women included in the analysis (median age 34 years, interquartile range 31–37), the within-individual risk of admission for psychosis in the 6-month postpartum period was 0.56 times (RRR, 95% CI 0.36–0.87) lower in continuers (IRR = 1.31, 95% CI 0.89–1.72) than in discontinuers (IRR = 2.34, 95% CI 1.87–2.91). Among discontinuers, the IRRs of admission for psychosis in the 6-month postpartum period did not change significantly with the timing of discontinuation (trend P = 0.946).

Antipsychotic continuation during pregnancy was associated with a reduced risk of postpartum relapse in women with schizophrenia. Continuing antipsychotics during pregnancy would be recommended after a risk–benefit assessment.

Accumulating evidence suggests that alterations in inflammatory biomarkers are important in depression. However, previous meta-analyses disagree on these associations, and errors in data extraction may account for these discrepancies.

PubMed/MEDLINE, Embase, PsycINFO, and the Cochrane Library were searched from database inception to 14 January 2020. Meta-analyses of observational studies examining the association between depression and levels of tumor necrosis factor-α (TNF-α), interleukin 1-β (IL-1β), interleukin-6 (IL-6), and C-reactive protein (CRP) were eligible. Errors were classified as follows: incorrect sample sizes, incorrectly used standard deviation, incorrect participant inclusion, calculation error, or analysis with insufficient data. We determined their impact on the results after correction thereof.

Errors were noted in 14 of the 15 meta-analyses included. Across 521 primary studies, 118 (22.6%) showed the following errors: incorrect sample sizes (20 studies, 16.9%), incorrect use of standard deviation (35 studies, 29.7%), incorrect participant inclusion (7 studies, 5.9%), calculation errors (33 studies, 28.0%), and analysis with insufficient data (23 studies, 19.5%). After correcting these errors, 11 (29.7%) out of 37 pooled effect sizes changed by a magnitude of more than 0.1, ranging from 0.11 to 1.15. The updated meta-analyses showed that elevated levels of TNF- α, IL-6, CRP, but not IL-1β, are associated with depression.

These findings show that data extraction errors in meta-analyses can impact findings. Efforts to reduce such errors are important in studies of the association between depression and peripheral inflammatory biomarkers, for which high heterogeneity and conflicting results have been continuously reported.