DISCUSSION
In this prospective study with medication diaries as the reference standard, we observed that majority of medications dispensed were reported to be used, ranging from 69% (medication for occasional and short-term use) to 88% (medication for chronic conditions). A substantial proportion of women, however, did not take the medication as prescribed: the overall implementation adherence was 74.2%. Achieving an implementation adherence of ≥80% ranged between 19% for medication for occasional and short-term use and 70% for medication for chronic conditions. Implementation adherence of pregnancy-related medication seemed lower for women aged >30 years, women with an income ≥\euro3,500, and non-smoking women, whereas women who smoked during pregnancy seemed to have a lower implementation for medication for occasional and short-term use.
Self-report based on interviews or questionnaires is generally acknowledged to result in underreporting of medication use during pregnancy [17-22]. This may explain why the proportion of medications dispensed that is actually reported to be taken is higher in the current study compared with previous studies, in which this measure ranged between 43 and 57% [14, 19]. However, consistent patterns of actual use are observed between studies when focusing on individual medications or groups of medications, with high levels of actual use for medication for chronic conditions, and substantially lower for medication for occasional and short-term use [14, 15, 19-21, 31-34]. Part of these lower levels of actual use for medication for occasional and short-term use may be explained by forgetfulness when using interviews or questionnaires that cover a prolonged period of pregnancy [22], but remained when using diaries as reference standard [21]. This disagreement may be worrisome for some maternal conditions, including infections (antibiotics) and fever (antipyretics), in which pharmacological treatment is often recommended to prevent maternal and infant morbidity and mortality [35-38]. In an Italian cross-sectional study, 3.2% of women reported intentionally avoiding a prescribed antibiotic in pregnancy [39].
Although actual use gives an indication of medication adherence among pregnant women, it does not assess whether medication is actually used as prescribed in terms of timing and dosage. As human fetal development is strictly sequenced with specific organ systems being formed at particular stages of pregnancy, timing of exposure is of utmost importance when assessing medication safety in pregnancy. In the current study, 42% of medications were not initiated on the date of prescription fill, but 1-7 days afterwards. These relatively small differences may result in exposure misclassification for the etiologically relevant time window when relying on prescription databases.
In our study, suboptimal implementation adherence could result from two distinct deviations of prescription instructions: an incorrect dosage and a shorter duration of use (persistence). Comparable to our study, de Jonge et al. [16] observed that a substantial proportion of women (30%) did not adhere to treatment recommendations, often using a lower dose or for a shorter duration than prescribed. Inadequate treatment of infections and pregnancy-related and chronic conditions, such as hypothyroidism [40], asthma [41], and diabetes [42], is associated with increased risk of adverse birth outcomes. Therefore, adequate and continuous dosage should be aimed for during pregnancy.
To date, there is no strong evidence for certain maternal and pregnancy characteristics to be associated with not following treatment recommendations. In addition to the associations with maternal age, household income, and smoking observed in this study, others identified associations with educational level, marital status, disease severity, and alcohol consumption [19-21, 31, 39, 43]. None of these findings have consistently been reported; only smoking status has been implicated in multiple studies [19, 20]. Identification of characteristics linked to non-adherence may facilitate the implementation of personalized preventive measures.
The clinical implications of non-adherence are relatively straightforward: it may endanger maternal and child health and increase healthcare costs [2, 3]. Enhancing adherence among pregnant women, however, is not straightforward. Improved counseling may attenuate some of the incorrect risk perceptions of medication safety among pregnant women, but for many medications, robust data on fetal safety are lacking [44-48]. This may hamper evidence-based decisions on whether the potential risks of pharmacological treatment outweigh the burden of the underlying maternal condition. Non-adherence also affects research on medication safety in pregnancy, as many studies rely on administrative databases for exposure assessment [23]. Not taking medication as prescribed, or not taking the medication at all, will lead to exposure misclassification and may bias studies determining associations with infant outcomes [49].
A major strength of the current study is the use of a diary to assess actual medication use, thereby reducing the probability of underreporting due to forgetfulness [24, 25]. The amount of details collected with these diaries enabled us to look beyond actual use by assessing initiation and implementation. Furthermore, the study population was considered large compared to some earlier studies, but still limited in the ability to assess adherence to individual medications and associations between maternal characteristics and implementation. Another limitation includes the relatively short time frame of only 6 weeks in mid-pregnancy. Adherence is expected to be lower in the first trimester of pregnancy because of fear for teratogenicity, but it was considered not feasible to administer diaries immediately after inclusion in the PRIDE Study until delivery due to participant burden. Mobile applications to daily record medication intake may improve the feasibility of intensive long-term data collection [50]. Social desirability might have increased reporting of medication use when not actually taken, but was considered unlikely as participants were unaware of the aim of this study. Also, we were unable to distinguish between intentional and unintentional avoidance of medication use. Some women may have discontinued treatment or implemented dose changes following consultation with a healthcare professional before the registered end date of the prescription, which may have decreased implementation adherence unjustly.
In conclusion, actual use of medication dispensed during pregnancy was high in this study compared with previous studies, but many women did not adhere to treatment recommendations regarding dosage and therapy duration. As this behavior may be detrimental to maternal and fetal health, future efforts should focus on identifying reasons for non-adherence and associated preventive measures. Furthermore, the results of this studies may inform and facilitate quantitative bias analyses in studies on medication safety in pregnancy relying on administrative databases or registries.