Stratification of the risk of bipolar disorder recurrences in pregnancy and postpartum

Arianna Di Florio, Katherine Gordon‐Smith, Liz Forty, Michael R. Kosorok, Christine Fraser, Amy Perry, Andrew Bethell, Nick Craddock, Lisa Jones, Ian JonesThe British Journal of Psychiatry 2018 · DOI

TL;DR

This study identifies previous perinatal psychiatric history as the most significant predictor of bipolar disorder recurrence in subsequent pregnancies. By stratifying risk based on prior episodes, clinicians can offer more personalized guidance to women with bipolar disorder considering pregnancy.

What Problem Does This Paper Address?

Women with bipolar disorder face a significantly elevated risk of mood episode recurrence during pregnancy and the postpartum period. This poses a complex dilemma: continuing mood-stabilizing medications may carry risks for the fetus, while discontinuing them increases the likelihood of a relapse. Existing clinical guidance often lacks the granularity to accurately assess individual risk, leading to difficult decisions for both patients and clinicians. Specifically, there's a need to understand how a woman's previous perinatal psychiatric history influences her risk in subsequent pregnancies, moving beyond general risk factors.

Methods

The researchers employed a retrospective study design, gathering data from 887 women with bipolar disorder who had at least two live births. Information was collected via semi-structured interviews, questionnaires, and case-note reviews.

Key analytical approaches included:

  • Backwards stepwise logistic regression: A standard statistical technique used to identify significant predictors of an outcome variable (in this case, perinatal recurrence).
  • Conditional permutation random forests: A machine learning algorithm that can handle complex interactions between variables and is robust to overfitting, used here to assess the importance of various clinical predictors.
  • Reinforcement learning trees: Another machine learning approach employed for variable selection, likely chosen for its ability to model non-linear relationships.

The study focused on two primary outcome variables: perinatal affective psychosis (for women with bipolar I or schizoaffective disorder, bipolar type) and perinatal non-psychotic depression (for the entire sample).

Datasets

The study utilized data collected from 887 women with bipolar disorder who were part of the Bipolar Disorder Research Network project. This dataset was enriched with retrospective information on their psychiatric history, obstetric history, and details of their first and second pregnancies and postpartum periods. No new datasets were introduced.

Key Insights and Ideas

  • Past is Prologue (for Perinatal Episodes): The most crucial takeaway is that a woman's history of a perinatal affective psychosis or depression is the strongest predictor of recurrence in a subsequent pregnancy. This highlights the importance of detailed perinatal psychiatric history in risk assessment.
  • Beyond Severe Episodes: The study emphasizes that even non-psychotic perinatal depression significantly increases the risk of recurrence, not just for depression but also for affective psychosis. This challenges a sole focus on severe psychotic episodes and underscores the need to consider all types of perinatal mood disturbances.
  • Risk Exists Even Without Prior History: A significant finding is that even women with bipolar disorder who had no perinatal mood episodes in their first pregnancy still have a substantial risk (around 34%) of experiencing an episode in a subsequent pregnancy. This means "no history" doesn't equate to "no risk."
  • Predicting Presentation and Timing: The clinical presentation (affective psychosis vs. non-psychotic depression) and timing of onset of a first perinatal episode appear to influence the presentation and timing of subsequent episodes. This offers a more nuanced understanding for risk stratification.

Key Results

  • Women with a history of perinatal affective psychosis were six times more likely to experience a recurrence of affective psychosis in their second pregnancy (OR = 6.0, 95% CI 3.43-10.90).
  • A history of non-psychotic depression in the first perinatal period significantly increased the odds of a non-psychotic depressive recurrence in the second pregnancy (OR = 3.3, 95% CI 2.10-5.23).
  • Overall, 55% of women with any previous perinatal psychiatric history experienced a recurrence in their second pregnancy, compared to 31% of those with no prior perinatal history.
  • For women with bipolar I disorder or schizoaffective disorder, bipolar type, the rate of any recurrence in the second perinatal period was 55% for those with a previous episode and 34% for those without.
  • For women with bipolar II disorder or bipolar disorder not otherwise specified, the recurrence rate was 56% for those with a previous episode and 27% for those without.
  • Onset of affective psychosis was almost exclusively within the first 6 weeks postpartum (90%), while non-psychotic depression had a more spread-out onset (59% in the first 6 weeks).

Limitations and Future Work

The authors acknowledge several limitations:

  • Retrospective Design: The study relies on retrospective data, meaning diagnoses were allocated based on the best estimate at the time of assessment, not necessarily specific to each pregnancy.
  • Medication Data Complexity: While some medication data was collected, the complexity of prescription regimes, adherence, and pharmacokinetic changes during pregnancy made it difficult to draw definitive conclusions about medication effects.
  • Exclusion of Hypomania: Hypomanic episodes were not included in the analyses due to challenges in retrospective assessment and defining clinical relevance.
  • Potential Unmeasured Variables: The authors note that factors like mood stability during pregnancy, complex longitudinal illness patterns, and medication adherence were not fully assessed and could influence outcomes.

Future work could involve prospective studies to confirm these findings and explore the impact of specific interventions and medications more rigorously. The development of a risk assessment flowchart (Fig. 2 in the paper) is a direct step towards applying these findings in clinical practice.

Summary updated

Abstract

Previous perinatal history of affective psychosis or depression is the most important predictor of perinatal recurrence in women with bipolar disorder and can be used to individualise risk assessments.Declaration of interestNone.
BibTeX
@article{florio2018stratification,
  author = {Arianna Di Florio and Katherine Gordon‐Smith and Liz Forty and Michael R. Kosorok and Christine Fraser and Amy Perry and Andrew Bethell and Nick Craddock and Lisa Jones and Ian Jones},
  title = {{Stratification of the risk of bipolar disorder recurrences in pregnancy and postpartum}},
  journal = {The British Journal of Psychiatry},
  year = {2018},
  doi = {10.1192/bjp.2018.92},
}
Figure