This study demonstrated the quality and utility of the JDIIP database for PregPV. Our assessment revealed that a substantial proportion of the ConcePTION CDEs in the framework established by the ConcePTION Project were either directly collected or derivable from existing variables within the JDIIP database. This strong alignment with established standards underscored the quality and comprehensiveness of the data captured during the consultation process.

The JDIIP database showed an 80% follow-up rate for pregnancy outcomes, which is noteworthy for PregPV. A previous questionnaire survey of pharmaceutical companies in Japan revealed that they struggled with data collection on drug exposure during pregnancy, with 60% of companies reporting that they could collect data for less than 25% of all followed-up cases [28]. This indicates that, under the current voluntary reporting system for adverse events in Japan, obtaining comprehensive data on pregnant women is challenging. Additionally, cases without abnormal outcomes are often underreported in voluntary reporting systems [10, 29]. However, the JDIIP database uniquely facilitates data collection for all consultation cases with confirmed drug exposure, including those without abnormal outcomes, thus enabling a comprehensive examination of patient characteristics and real-world drug use patterns.

The JDIIP database was highly effective in capturing details related to database management, pregnancy specificity, maternal medical history, medication exposure during pregnancy, live or stillborn birth outcomes, and malformations (Table 2). These categories exhibited over 80% coverage of the necessary variables, highlighting the strength of the database in capturing critical information relevant to PregPV. However, certain areas, such as maternal medical conditions arising during pregnancy and infant complications within the first year of life, showed less alignment with the CDE framework. This discrepancy can be attributed to the nature of data collection, which primarily relied on spontaneous return visits by patients and postcard surveys conducted approximately 1 month after the expected delivery date. Consequently, dynamic changes in maternal conditions or long-term infant outcomes may have been underreported. To address these data gaps, enhancements to the JDIIP data collection are being considered. For example, data on infant outcomes up to 1 year of age have recently been added to the follow-up protocol, and this will help capture infant complications within the first year of life. Furthermore, integrating the JDIIP database with hospital electronic medical records could supplement currently missing clinical details (such as prenatal test results, specific drug indications, maternal death, or ectopic pregnancy) that are not collected in the current consultation-based system. In the future, linkage with structured public health data sources related to maternal and child health could also be explored to improve the completeness of long-term follow-up, although such integration is not yet in place. For data elements where the JDIIP provides information that differs from standard definitions (the “divergent” CDE category), efforts to harmonize these variables with internationally accepted definitions (e.g., the ConcePTION core data set) would enhance the comparability and utility of the data.

Despite these limitations, our descriptive analysis demonstrated the ability of the database to provide valuable insights into real-world scenarios. Key maternal characteristics, including age, pregnancy planning, medical history, and lifestyle factors, are well documented in the JDIIP database. Furthermore, this database offers comprehensive details on medication exposure during pregnancy, enabling the identification of frequently used drug classes and the timing of exposure. Notably, some variables with high clinical relevance, such as gestational age at the end of pregnancy and congenital abnormalities, were successfully collected from the database (Table 4). This is noteworthy because these data are not available in ICH E2B(R3), the standardized reporting format for spontaneous adverse event reporting systems [27], and so the JDIIP database potentially enables a more sophisticated risk assessment. The comprehensive nature of the database could represent the basis for facilitating risk evaluation by stakeholders, including marketing authorization holders and regulatory authorities, as well as primary reporters for pregnancy exposure reports.

Nonetheless, it is important to acknowledge the inherent limitations of the JDIIP database. First, as a consultation-based database, the JDIIP database may be subject to selection bias, because only women who seek consultation are included. Women who do not use any medications during pregnancy generally have less incentive to seek consultation; even if they do, their pregnancy outcomes are not followed up in the current system. As a result, the JDIIP cohort does not include an internal control group of unexposed pregnancies. Including follow-up data from pregnant women who did not use medications could serve as a valuable internal control group, thereby improving the validity of comparative risk assessments. However, expanding the follow-up system to include non-medication users would require additional procedural steps, including obtaining informed consent and securing ethical approval. Moreover, substantial logistical and human resources would be required to support follow-up of patients who are not part of the follow-up system. These considerations highlight the logistical and ethical challenges that must be addressed before expanding the follow-up system. It is also noteworthy that approximately one-third of participants in the JDIIP database were reported to have psychiatric disorders, including depressive and anxiety disorders. This relatively high prevalence may partly reflect our broad definition of psychiatric disorders, which encompasses a wide spectrum of conditions, and partly the characteristics of women more likely to seek consultation—particularly those undergoing psychotropic treatment and concerned about medication use during pregnancy. Some of these women may not have had severe underlying conditions, but rather sought reassurance due to heightened anxiety or uncertainty. In such cases, access to individualized consultation may have contributed to reduced anxiety and the successful continuation of the pregnancy, ultimately leading to healthy birth outcomes. This supportive role of the consultation service should be recognized as part of its broader contribution to PregPV. Nonetheless, such overrepresentation should be considered when interpreting findings related to psychiatric conditions or psychotropic drug exposure. Second, the follow-up period is relatively short, primarily focusing on pregnancy outcomes and early infancy, which limits the ability to assess long-term effects on child development. Third, the database may not capture all relevant confounding factors, such as genetic factors that could influence pregnancy outcomes, potentially limiting the ability to establish causal relationships between drug exposures and observed outcomes. Fourth, due to the absence of a nationwide registry linking pregnancy and birth outcomes in Japan, it is not currently possible to validate JDIIP data at the individual level using patient or birth records. To evaluate the representativeness of the JDIIP population in terms of key maternal and pregnancy characteristics, we compared selected JDIIP variables with recent national statistics. The median maternal age at childbirth in the JDIIP cohort was 32.0 years, comparable to the national average of 32.2 years in 2022, steadily rising from 31.9 years in 2018 [30]. The preterm birth rate (< 37 weeks) among singleton live births in the JDIIP dataset was 7.2%, which aligns with national estimates of approximately 5–6%. These findings suggest that the JDIIP cohort demonstrates reasonable epidemiological representativeness for use in PregPV research. Finally, since no patient residence data (e.g., region or urban/rural status) are collected, we could not directly evaluate regional representativeness. Nonetheless, the JDIIP consultation network spans all 47 prefectures of Japan, likely capturing a geographically diverse sample and thereby reducing geographic selection bias.

To fully leverage the JDIIP database in assessing the effects of drug exposure on pregnant women and fetal outcomes, further accumulation of cases is needed. Specifically, the relatively small sample size might make it challenging to explore rare events adequately. Additionally, the data acquisition timing was skewed toward when consultations occurred, primarily during early pregnancy, potentially affecting the representation of various pregnancy stages, and later pregnancy exposures or outcomes. Gestational age at the time of consultation is an important contextual factor that affects the scope and reliability of data captured. As previously reported [24], the median gestational age at the time of consultation was 10 weeks. This indicates that most consultations occurred during the first trimester. Accordingly, drug exposure data during early pregnancy are well represented, whereas exposures or complications arising later in pregnancy may be underreported, particularly among those who did not return for follow-up. Recent system improvements have enabled the collection of updated information on medication changes during mid to late pregnancy to mitigate this limitation. Although restricting analyses to women who sought consultation after week 12 might reduce uncertainty regarding first-trimester exposures in some cases, such an approach may also introduce greater recall bias due to the longer time lapse between exposure and reporting. Moreover, it would limit the generalizability of the findings in early pregnancy safety studies, which are often the primary focus of PregPV. Therefore, we retained all cases in the present analysis and acknowledged this limitation. Future studies may benefit from sensitivity analyses stratified by gestational age at consultation.

Another important consideration in evaluating the completeness of pregnancy outcome data is the observed miscarriage rate. The miscarriage rate observed in the JDIIP database (7.4%) appears lower than the commonly reported rate of approximately 10–15% in clinically recognized pregnancies internationally [31]. In Japan, a recent analysis of the Japan Environment and Children’s Study cohort indicated that 15.3% of pregnant women had a history of miscarriage or stillbirth [32]. This discrepancy likely reflects the timing and structure of data capture in the JDIIP system. The median gestational age at the time of consultation is 10 weeks [24], whereas most miscarriages occur in the early first trimester, particularly before 10 weeks of gestation [33]. As such, early pregnancy losses may not be captured if consultation has not yet occurred. In addition, women may be more likely to seek consultation when the pregnancy is ongoing, leading to underrepresentation of early losses. These factors likely contribute to the lower miscarriage rate observed in this cohort. These limitations underscore the need to interpret the results carefully and to recognize the inherent constraints of the data. However, collecting all necessary information within a single database system poses a challenge. Therefore, obtaining the required data through appropriate collaboration with multiple sources should be considered [34]. From a collaborative perspective, the current database, which primarily relies on voluntary patient consultations, may not include information on drugs with high data evaluation demands, such as newly marketed drugs or those associated with certain safety concerns. Promoting awareness and recognition of the database registration system through collaboration with healthcare professionals and academic societies is essential for addressing this issue. Considering the potential use of the database for PregPV, pharmaceutical companies, and industry groups could also help promote awareness of these data within the medical community. From a regulatory perspective, authorities could consider conducting safety evaluations and implementing measures, such as revising package inserts using data from the JDIIP database. For pharmaceutical companies, effective use of the JDIIP database may help overcome data collection challenges through an adverse event spontaneous reporting system, which could be integrated with routine pharmacovigilance activities or serve as a partial alternative approach, particularly in response to specific safety concerns. Sustainable data collection strategies to achieve larger sample sizes, diverse data collection timeframes, and extended observation periods may help mitigate the current limitations of PregPV and provide a more comprehensive understanding of the effects of drug exposure in pregnant women and their offspring.

This study had some limitations. First, the latest version of the CDE was used to ensure the objectivity of the database’s value assessment. However, the CDE is scheduled for continuous review, and parts of the initial version included in this study may differ from the latest online version. Second, data utility in the JDIIP database was evaluated based on the presence of relevant items and whether the data format was analyzable. The actual analyzability, including the adequacy of the data volume, depends on the specific analysis objectives.