In the present study, we found that use of prescribed drugs during pregnancy is prevalent in Sweden with a large variation during pregnancy and lactation periods. The agreement between drug use registered in the MBR and filled prescriptions recorded in the PDR was high for drugs used for chronic conditions, but not for drugs used more occasionally. As the information in the MBR is based on self-reported drug use and recorded in the antenatal medical records and the information in the PDR reflects filled prescriptions, the low agreement for occasionally used drugs between the two registers may reflect both nonreporting and noncompliance. Consequently, for occasionally used drugs, neither of the registers is more advantageous than the other as the only source of information for drug exposure, and for future studies, presumably the combined information will be the best estimate.
In the study by Olesen et al based on telephone interviews during pregnancy, agreement was high for drugs used for chronic diseases like beta-blockers, insulin, and thyroid therapy, which is similar to our findings.9
Previous studies have found that low socioeconomic status is associated with use of prescription medication during pregnancy, and as individuals with low socioeconomic status may be less prone to participate in surveys, the results from the Danish study may not be applicable to pregnant women in general or women with low socioeconomic status in particular.4
When information on drug use is obtained from interviews in surveys or in the clinical setting, the type of drug might influence recall accuracy. For drugs used more occasionally, there is a possibility that a woman fails to report drug use or prefers not reporting it to the midwife because she considers it to be of no importance or because she might fear that reporting the drug intake may complicate her pregnancy. A Dutch study concluded that interview data will outweigh data from pharmacy records for self-medication and compliance, whereas pharmacy records will give better information in case of long recall periods and in patients with multiple and repeated drug use.10
An additional and complicating aspect when estimating drug use from filled prescriptions is the time that might pass between drug purchase and medication start. Generally, in Sweden, a prescription will cover a period of 90 days. However, when we expanded the time period to include prescriptions filled up to 180 days prior to the first antenatal care visit, we found a better agreement between the information obtained from self-reported use and recorded in the MBR and the information on purchased drugs from the PDR, indicating a delay in starting a medication or using a lower dose than prescribed. We are not able to address whether this delay or altered dosing is the normal pattern of use of prescribed drugs among women in reproductive age or if it was the pregnancy per se that influenced drug use.
Drug use during pregnancy is complicated by the potential risk for both mother and fetus. For women with chronic inflammatory disease like rheumatoid arthritis and inflammatory bowel disease, use of medications like methotrexate is clearly teratogenic, whereas there is a lack of knowledge concerning the safety of TNF-alpha-blockers during pregnancy.17
We observed that corticosteroids for systemic use and immunosuppressant drugs were dispensed less during pregnancy compared to the 3-month period before pregnancy. As knowledge concerning adverse fetal effects of these drugs in humans is limited and serious adverse effects have been reported in animal studies, the precaution of reducing or avoiding systemic anti-inflammatory medication during pregnancy might be wise.18
On the other hand, unless pregnancy in itself reduces disease activity, stopping or reducing medication might increase risk of flares and worsening of the disease, which may also affect the fetus adversely. Agreement between self-reported and dispensed drugs was high for immunosuppressant therapy, whereas this was not found for systemic corticosteroid use. For thyroid therapy, on the other hand, use increased during pregnancy, and there was good agreement between self-reporting and expenditure data. These findings are in line with drug use during pregnancy in Denmark, with agreements of 100% and 20% for thyroid drugs and systemic corticosteroids, respectively.9
For studies on congenital abnormalities (CA), time of exposure is of importance because the period of teratogenesis may be quite short. There is a risk that studies based on linkage with prescription registers will be uncertain. This is less important for chronically used drugs compared with occasionally used drugs like antibiotics and sedatives.21
In addition, a teratogen does not uniformly increase the rates of all CA, but rather tends to increase rates of selected CA, which further emphasizes valid drug exposure information.22
Our finding of a low agreement between self-reported drug used as recorded in the MBR and purchased drugs recorded in the PDR for occasionally used drugs emphasizes the difficulties in using such information when studying these drugs in association with rare outcomes such as CA.23
Antibiotics were the most dispensed prescribed drug during pregnancy, and the proportion of women filling prescriptions with antibiotics increased further during the lactating period reaching almost 14%. The most prescribed antibiotic drugs were B-lactam antibacterials and penicillins, and we observed strong decline in dispensed antibiotics not recommended during pregnancy. Similar findings were reported in a Norwegian5
and a German25
study. Among women with a dispensed antibiotic drug from 90 days before first visit to antenatal care, only 20% reported this intake according to the medical records. Conversely, among women who reported use of antibiotics, 83% were dispensed according to the register within 90 days from first attendance at antenatal care.
The use of drugs for the nervous system decreased during pregnancy except for antipsychotics (N05A), where there was a strong increase in dispensed drugs during first trimester. Similar findings were reported from Norway.5
The safety of antipsychotics during pregnancy remains unclear, but women who discontinue treatment during pregnancy have an increased risk of relapse.26
Antipsychotic use during pregnancy has been associated with increased risk of CA27
and preterm delivery, whereas risks of low birth weight and small-for-gestational-age birth were not influenced by antipsychotic use in a large Taiwan study.26
A major strength of the present study is the use of a population-based cohort comprising almost all deliveries in Sweden in 2007. Data from the PDR contains all expenditures of prescribed drugs in pharmacies in Sweden. Hence, we have an almost complete coverage of prescribed drugs dispensed during pregnancy. Information in the MBR on drug use and maternal characteristics is prospectively reported in early pregnancy, and consequently, the risk of recall bias is limited.
A limitation of this study is the lack of information on drugs used in hospitals. Hence, there may be an underreporting of drugs administrated by infusion and drugs for chronic psychiatric and rheumatic disease. However, this would most likely influence the findings for chronic conditions and not for occasional use. Furthermore, we do not have any information in the PDR on OTC drugs, and the low compliance between self-reported and register information on, for example, NSAID and analgetics may stem from OTC drugs or from drugs dispensed by the partner or other bound to share medication. In the present study, we included only women with a delivery from gestational week 22 and onward, and consequently, we were not able to study drug use in relation to miscarriage or terminations of pregnancies because of suspected malformations or drug use in early pregnancy. In addition, we chose to include only women giving birth to singletons, as pregnancies with more than one fetus, in general, are under closer surveillance with higher risks of drug treatment and women giving birth to more than one infant will be less prone to breast-feed, which might affect their drug intake. Accordingly, the results from this study reflect only women giving birth to singletons.