We undertook a register-based study using data from the Swedish Medical Birth Register (MBR), the Swedish National Patient Register (NPR), the Swedish National Prescribed Drug Register (NPDR) and Statistics Sweden. Since 1973, the Swedish MBR has been prospectively collecting information on all births in Sweden. This is facilitated by recording data from the mother´s first antenatal visit, until mother and child leave the hospital18. Data is usually recorded by midwives and physicians at clinical visits, then reported electronically by using a check-box system. Data on the mother´s medical history from in-patient care and specialist out-patient care is reported using the International Statistical Classification of Diseases and Related Health Problems 10th revision (ICD-10) codes collected from the NPR. Additionally, the NPDR collects information on all medications dispensed by Swedish pharmacies. We collected information on prescribed and filled prescriptions for antiviral medications and aspirin (routinely prescribed at low dosages for women at high risk of developing preeclampsia). Information on country of birth and years of education was collected from Statistics Sweden. Women were linked amongst registers through their personal identification number, which is assigned at birth or when achieving residency in Sweden. For all statistical analyses, we assumed compliance with the instructions for dispensed prescriptions. This study was approved by the Swedish Ethical Review Authority (approval number 2019–04925 with amendment number 2022 − 00922). According to Swedish legislation, informed consent from patients is not required for register-based studies. Access to data from these registers requires a request to the relevant government agency, following submission of a valid ethical review and approval. Consequently, in the context of our study, a statement on informed consent is not applicable, as Swedish law deems ethical approval to provide sufficient protection for participants in register-based research.
Study population
Data was collected on all women giving birth to their first child in Sweden between 2007 and 2019. Pregnancies that ended in stillbirth after the 22nd gestational week were included. Twin pregnancies were excluded.
Gestational length at delivery was calculated based on a routine antenatal ultrasound for 91.4% of pregnancies, based on estimated gestational length after delivery for 4.3% of women, and based on last menstruation supported by estimated gestational length for 3.8% of women.
Maternal characteristics included: age at delivery, body mass index (BMI; calculated from height and weight at first antenatal visit), country of birth (classified as Sweden, other Nordic countries, North America and Europe, or other countries). Housing situation was defined as living with a partner, living in a single household, or other situation. Level of education was recorded as having less than 12 years of schooling (not completed high school), having completed high school, or having completed university. Information on smoking status, use of in vitro fertilization (IVF) and pre-gestational disorders (chronic hypertension, kidney disease, systemic lupus erythematosus, diabetes) was collected through MBR (using pre-defined checkboxes) and NPR (ICD-10 codes: pre-gestational diabetes: O240, O241).
Information on herpes simplex diagnosis during pregnancy (ICD-10 codes: A60, B00) was also collected as it is the primary reason for antiviral medication prescription during this period.
Exposure
We defined exposure as filling an antiviral medication prescription, specifically for nucleosides and nucleotides, excluding reverse transcriptase inhibitors with Anatomical Therapeutic Chemical (ATC) classification code J05AB, during pregnancy or 90 days before estimated conception date. Antiviral medications prescribed in Sweden with ATC-code J05AB include: Aciclovir, Ganciclovir, Famciclovir, Valaciclovir, Cidofovir, Valganciclovir and Remdesivir. To investigate the significance of timepoint in antiviral medication use, women filling a prescription for antiviral medications were further divided into three categories based on when the first prescription was filled: before pregnancy (90 days before conception) or first trimester (up to week 12), second trimester (weeks 13 to 27), and third trimester (weeks 28 to delivery). The use of antiviral medications with ATC-code J05AB is considered safe during pregnancy19,20,21 and their use is not associated with an increase in birth defects22.
Outcomes
Preeclampsia was the primary outcome, recorded as ICD-10 codes: O11, O14, O15 and O16 in MBR and NPR. A distinction was made between different preeclampsia outcomes, depending on gestational age at delivery or infant growth. The following analyses were repeated for each: preeclampsia with early preterm delivery defined as preeclampsia with delivery before week 34 (preeclampsia < 34 weeks), preterm preeclampsia defined as preeclampsia with delivery before week 37 (preeclampsia < 37 weeks), term preeclampsia defined as preeclampsia with delivery after week 37 (preeclampsia ≥ 37 weeks), and preeclampsia with a small-for-gestational-age infant (preeclampsia with SGA). SGA was defined as a birthweight below two standard deviations of mean birthweight for gestational age and fetal sex23.
Statistical analysis
Population characteristics were presented according to antiviral medication use. Women taking antiviral medication were further described according to time point of first dispensed prescription. Results are presented as unadjusted and adjusted odds ratios, with a 95% confidence interval (OR or aOR with 95% CI).
As a form of sensitivity analysis, the group unexposed to antiviral medications was further divided according to HSV status. Differences in likelihood of developing preeclampsia between the three groups (I. Women unexposed to antivirals, HSV-, II. Women unexposed to antivirals, HSV + and III. Women exposed to antivirals) were explored using logistic regression.
All statistical analyses were performed using R Statistical Software (RStudio 2023.09.1 + 494 “Desert Sunflower”). P-values below 0.05 were considered statistically significant.
Confounders
Possible confounders were evaluated by making a directed acyclic graph (DAG), considering clinical risk factor for preeclampsia from the existing literature (Supplementary Fig. 1). The confounders identified were: maternal age, BMI, country of birth, IVF use, smoking at first antenatal visit, pre-gestational diabetes, chronic hypertension, chronic kidney disease and systemic lupus erythematosus.
Inverse probability of treatment weighting
The association between use of antiviral medications and the development of preeclampsia was explored using logistic regression with inverse probability of treatment weighting (IPTW). For this analysis, the control group included all women unexposed to antiviral medications. Briefly, IPTW is a method that can be used to adjust baseline characteristics of a population in an observational study24,25. IPTW was done in two steps. First, the propensity score for average treatment effect (ATE) was calculated using the exposure as the outcome and weights were stabilized with the WeightIt package26 in R. Covariate balance was assessed graphically using the cobalt27 R package. aOR with 95% CI were calculated using robust standard errors, computed with the R package clubSandwich.28 Three different imputations were made for missing BMI values and the analysis was repeated three times for each imputed BMI dataset. Imputations were made using the mice29 R package with predictive mean matching (pmm), using our confounders (as identified in the DAG) as predictors.
Model fit and aORs were calculated for each different BMI imputation, with results pooled together using Rubin´s combining rule30. For time of exposure subgroup analysis, a new propensity score was calculated for each of the three timepoints. Again, model fit and odds ratios were calculated for each different BMI imputation, with results pooled together.
For all analyses involving different preeclampsia outcomes, eight individuals were excluded because of unknown gestational age.
