March 22, 2012

antidepressants and hypertension in pregnancy

This analysis accompanied a roundup which can be viewed here.

Title, Date of Publication & Journal

‘Antidepressant Use During Pregnancy and the Risk of Pregnancy Induced Hypertension’ by De Vera, M et al. published in British Journal of Clinical Pharmacology on 15th March 2012.

Claims supported by evidence?

• The analysis suggests increased hypertension when using antidepressants, but only weakly. Some results are borderline significant, but that could be because of multiple testing of different types of antidepressant, as the authors state later in the paper. 
• The study does not demonstrate sufficiently that the risk of hypertension from anti-depressants during pregnancy is above the risk of hypertension from depression during pregnancy; or that one class of drugs or one specific drug is largely responsible for the risk.

Summary

• Study based on a large number of patients, but results only borderline significant
• Evidence is sufficient to initiate a new prospective study, but not sufficient to recommend changes to current medical guidelines for women
• Study carried out multiple tests for different drugs, without adjusting sufficiently for this methodology 
• Interpretation focused on just one cause (antidepressants) – other medications could be responsible as indicated by the study data.

Study Conclusions

The paper does suggest an effect of a class of drugs on hypertension during pregnancy. But it is based on an exploratory analysis of a large database, not on a pre-planned prospective study, which weakens the evidence of a clear causal link. However it is an interesting indicator which should prompt further investigation.

The increase in incidence for the specific SSRI, paroxetine (by a reported 81%) is not demonstrated with statistical significance and its magnitude could be down to chance, as there were many tests performed without adjusting for the number of tests. That means the 81% could just be a chance outlier. The apparent significance of the effect for paroxetine is further impacted by the fact that the majority (>50%) of the patients were taking that particular drug, while all other drugs in this analysis were used much less – so their data are currently even more inconclusive.

The paper does not entirely support the hypothesis that the risk is “above that which may be attributed to depression or anxiety disorders in the absence of drug treatment”. This is because it only presents the risks attributed to depression before pregnancy, not during pregnancy. It is therefore possible, for example, that the hypertension could be attributed to depression alone.

Strengths/Limitations

Study design and interpretation

• (+) Case-control design is a typical design for such epidemiological studies
• (+) Performance of analyses according to standards
• (-) Interpretation of results does not account for the large number of tests that were performed– there may be outliers
• (-) The effects of anti-depressants on blood pressure in other populations (men and non-pregnant women) are only very briefly discussed. This would normally be anticipated, so the mechanism for the finding remains unclear
• (-) Use of anti-depressants is noted before and during pregnancy; but psychiatric diagnosis only made before pregnancy.

Controls

• (+) Large number of control subjects selected
• (-) Controls not matched for use of antidepressants before pregnancy 

Incidence in general population

• (-) Author statement: “In our study, fewer women continued antidepressants after the first trimester, limiting our ability to detect significant associations.”

Glossary

• Case-control study design: a very practical design, but the evidence is lower than for most other study types (e.g. randomised trials or prospective cohort studies)
• Matched controls: For each subject with pregnancy induced hypertension, 10 subjects without that condition are selected from the database using relevant criteria
• Multiplicity adjustment of analyses: The more statistical tests performed, the more likely it becomes that one test will show an extreme result – just because the data always contain some randomness, even if this is in fact wrong. This study performed many statistical tests.

 ‘Before the headlines’ is a service provided to the SMC by volunteer statisticians: members of the Royal Statistical Society (RSS), Statisticians in the Pharmaceutical Industry (PSI) and experienced statisticians in academia and research.  A list of contributors, including affiliations, is available here.