Selective serotonin reuptake inhibitors and venlafaxine in early pregnancy and risk of birth defects: population based cohort study and sibling design

I presented this article at a local journal club and it generated some interesting discussions.

A major component of this was how I had tried to explain the results. One way to make cohort study results easier to understand is to express the results as number needed to harm. So, first I constructed a 2x2 table:
Birth defect No defect
SSRI/Venlafaxine A
1357 B
35,415 (a+b)
36,772
Not exposed C
71,374 D
2,195,501 (c+d)
2,266,875
Totals 72,731
(a+c) 2,230,916
(b+d) 2,303,647

So:
• RR is probability of event occurring/total number of events possible
• = EER/CER
• = (a/a+b)
(c/c+d)
• = (1357/36,772)/ (71,374/2266875) = 1.18
• Absolute risk reduction or increase as well as number needed to treat/harm (NNT/NNH)
• ARR = CER – EER
= 0.0369 – 0.0314
= -0.0055 (negative value = increased risk, therefore risk of harm)
NNH = 1/ARR
1/0.0055
= 181

So 181 mothers need to take SSRI/Venlafaxine for 1 additional birth defect to occur

Our critical appraisal discussion concluded that overall there were several strengths of the study:
- It was a large, relatively well conducted study
- Mandatory health registers takes away several biases
- Very detailed study with lots of analyses based on previous evidence – but hard to understand in places

Some limitations in the study included:
- Lack of consideration for some confounding factors, including depression diagnosis alone
- Sibling analyses lacked statistical power

The overall results indicate that from this study, there was a very small absolute risk (NNH 181) of birth defects from taking SSRIs or Venlafaxine during pregnancy and overall a small relative increase compared to the control group, which some people found very reassuring whilst others considered it a risk too high when the stakes are as important as birth defects.