Intended for healthcare professionals

CCBYNC Open access

Rapid response to:

Research

Selective serotonin reuptake inhibitors during pregnancy and risk of persistent pulmonary hypertension in the newborn: population based cohort study from the five Nordic countries

BMJ 2012; 344 doi: https://doi.org/10.1136/bmj.d8012 (Published 12 January 2012) Cite this as: BMJ 2012;344:d8012

Rapid Response:

Re: Selective serotonin reuptake inhibitors during pregnancy and risk of persistent pulmonary hypertension in the newborn: population based cohort study from the five Nordic countries

Responses to Professor Bewley, Principal Research Associate Petersen and Dr Newcomb.

We thank Professor Bewley for the interest in our paper.(1) The aim of our study was to assess an association between SSRI and PPHN and other possible risk factors for PPHN were therefore not assessed. As Professor Bewley points out the infants included in our study and exposed to SSRIs were more often born preterm (5.4% vs. 3.8%) and by cesarean section (19.0% vs. 13.7%) when compared to the unexposed. In our risk analyses we did adjust for several possible confounders. However, neither gestational age at birth or mode of delivery should be considered confounders, as they could not have influenced the exposure to SSRI, and consequently they were not included in the multivariate analyses. A previous report suggested that the high risk of PPHN found in association with SSRI use in late pregnancy could be explained by cesarean delivery rather than SSRI exposure.(2) As described in our paper, we carried out interaction analyses to test whether mode of delivery modified the effect of exposure to SSRIs and found no such effect. In addition we performed analyses to assess risks of PPHN, when restricting the population to infants born gestational week 37 to 41 and delivered vaginally and found an OR of 2.5 (95% CI: 1.5-4.2) for SSRI exposure in late pregnancy. These additional analyses were not reported in the paper.

We thank Principal Research Associate Petersen and colleagues for their comments to our paper and for raising the issue of confounding by severity. They suggest adjusting for severity of disease by including previous psychiatric disease as an independent risk factor in the analyses. However, we are not convinced that such analyses would be superior to our stratified analyses by previous hospitalization for a psychiatric disease and exposure to SSRIs. Petersen and colleagues also suggest that women prescribed SSRI in late pregnancy might differ from those prescribed their drugs earlier in pregnancy, which they did, but only to a small extent. Those exposed in late pregnancy were older (22.6% were 35 years of age or older vs. 20.6% in early pregnancy), the infants were more often born preterm (6.6% vs. 4.5%) and more often delivered by cesarean section (20.7% vs. 17.7%). Otherwise the two exposure groups were rather similar.

We identified possible confounders in advance such as BMI and smoking. However, as the cohort included data from the five Nordic countries during a ten year period with varying coverage on certain variables, we had to reconsider our initial intention to include all the predefined confounders, to avoid losing important information. As Norway had almost 20% missing information on smoking and Iceland did not include this information in their birth register we tested whether including smoking in the model on data from Denmark, Finland and Sweden (89% of the population) affected the risk estimate and found no such effect. We used a similar methodology for BMI and found no effect on the risk estimate in the subcohort. Accordingly, we omitted smoking and BMI in the final model. Also Petersen and colleagues suggest using propensity scores instead of adjusting for possible confounders in multivariate regression analyses. We agree that propensity scores sometimes might be useful, in particular when detailed information is available. However, though the information in the Nordic registers is rich in many ways the registers include only certain key variables. As analyses using propensity scores or multivariate modeling usually generate similar results when including a limited number of factors we chose the multivariate modeling. (3) We also want to point out that we did include country and level of hospital in the adjusted analyses. Petersen and colleagues suggest including only women who have been dispensed two or more SSRIs within a short time interval prior to delivery. We also agree that being dispensed more than one prescription should strengthen the assumption that a filled prescription equals drug intake. However, we do not consider such approach meaningful for this study considering that exposure in late pregnancy covered 20 weeks and prescriptions generally are filled for 12 weeks. Women who filled prescriptions with shorter intervals than 12 weeks presumably differed from those who did not, as shorter filling intervals might imply worsening of the underlying disease. Though a prescription could be filled in late pregnancy to enable start of medication directly after delivery, we do not find this likely for antidepressants. Considering the mood changes, which occur in connection with delivery it is difficult to foresee the need for medication after delivery.

We are happy to sort out the uncertainties raised by Dr Newcomb. The exposure periods were set in advance based on the previous report by Chambers et al with increased risks for PPHN in association with exposure in gestational week 20 or later.(4) As a prescription covers 12 weeks, the exposure period in early pregnancy was set to 8 weeks. We only had information on prescribed drugs and cannot exclude that some women might have bought NSAIDs over the counter. Designing studies is often a trade of between sample size and detailed information. As a very large study was needed to test our hypothesis we had to compromise on the information level. We also want to point out that our data are not retrospective. All data included in the Nordic registers are collected prospectively, i.e. exposure information on filled drugs, maternal characteristics including previous hospitalization for psychiatric disease was collected before outcome of the study. For questions regarding smoking and BMI please see previous section.

References
1. Kieler H, Artama M, Engeland A, Ericsson O, Furu K, Gissler M, et al. Selective serotonin reuptake inhibitors during pregnancy and risk of persistent pulmonary hypertension in the newborn: population based cohort study from the five Nordic countries. BMJ. 2012;344:d8012. Epub 2012/01/14.
2. Wilson KL, Zelig CM, Harvey JP, Cunningham BS, Dolinsky BM, Napolitano PG. Persistent pulmonary hypertension of the newborn is associated with mode of delivery and not with maternal use of selective serotonin reuptake inhibitors. Am J Perinatol. 2011;28(1):19-24. Epub 2010/07/08.
3. Sturmer T, Joshi M, Glynn RJ, Avorn J, Rothman KJ, Schneeweiss S. A review of the application of propensity score methods yielded increasing use, advantages in specific settings, but not substantially different estimates compared with conventional multivariable methods. Journal of clinical epidemiology. 2006;59(5):437-47. Epub 2006/04/25.
4. Chambers CD, Hernandez-Diaz S, Van Marter LJ, Werler MM, Louik C, Jones KL, et al. Selective serotonin-reuptake inhibitors and risk of persistent pulmonary hypertension of the newborn. N Engl J Med. 2006;354(6):579-87.

Competing interests: No competing interests

17 March 2012
Helle Kieler
Head of Centre for Pharmacoepidemiology
Miia Artama, Anders Engeland, Örjan Ericsson, Kari Furu, Mika Gissler, Rikke Beck Nielsen, Mette Norgaard, Olof Stephansson, Unnur Valdimarsdottir, Helga Zoega, Bengt Haglund
Karolinska Institutet
CPE, T2, Karolinska University Hospital , 17176, Stockholm, Sweden