Drug Safety (2023) 46:689–702 https://doi.org/10.1007/s40264-023-01314-3 ORIGINAL RESEARCH ARTICLE Impact of 2018 EU Risk Minimisation Measures and Revised Pregnancy Prevention Programme on Utilisation and Prescribing Trends of Medicinal Products Containing Valproate: An Interrupted Time Series Study Shahab Abtahi1 · Romin Pajouheshnia1 · Carlos E. Durán2 · Judit Riera‑Arnau2,3 · Magdalena Gamba1 · Ema Alsina2 · Vjola Hoxhaj2 · Morten Andersen4 · Claudia Bartolini5 · Sarah Brøgger Kristiansen4 · Jeremy Brown6 · Christine Erikstrup Hallgreen7 · Patricia Garcia‑Poza8 · Helga Gardarsdottir1,9 · Rosa Gini5 · Anna Girardi5 · Emily Holthuis10 · Consuelo Huerta11 · Luisa Ibánez3 · Giorgio Limoncella5 · Mar Martín‑Pérez8 · Olga Paoletti5 · Giuseppe Roberto5 · Patrick Souverein1 · Karin M. A. Swart10 · Kevin Wing6 · Miriam Sturkenboom2 · Olaf Klungel1 Accepted: 25 April 2023 / Published online: 9 June 2023 © The Author(s) 2023 Abstract Introduction Due to established teratogenicity of valproates, the EU risk minimisation measures (RMMs) with a pregnancy prevention programme (PPP) for valproate were updated in March 2018. Objectives To investigate the effectiveness of the 2018 EU RMMs on valproate utilisation in five European countries/regions. Methods A multi-database, times series study of females of childbearing potential (12–55 years) was conducted using elec- tronic medical records from five countries/regions (01.01.2010–31.12.2020): Denmark, Tuscany (Italy), Spain, the Nether- lands, and the UK. Clinical and demographic information from each database was transformed to the ConcePTION Common Data Model, quality checks were conducted and a distributed analysis was performed using common scripts. Incident and prevalent use of valproate, proportion of discontinuers and switchers to alternative medicine, frequency of contraception coverage during valproate use, and occurrence of pregnancies during valproate exposure were estimated per month. Inter- rupted time series analyses were conducted to estimate the level or trend change in the outcome measures. Results We included 69,533 valproate users from 9,699,371 females of childbearing potential from the five participating centres. A significant decline in prevalent use of valproates was observed in Tuscany, Italy (mean difference post-intervention −7.7%), Spain (−11.3%), and UK (−5.9%) and a non-significant decline in the Netherlands (−3.3%), but no decline in inci- dent use after the 2018 RMMs compared to the period before. The monthly proportion of compliant valproate prescriptions/ dispensings with a contraceptive coverage was low (<25%), with an increase after the 2018 RMMs only in the Netherlands (mean difference post-intervention 12%). There was no significant increase in switching rates from valproates to alterna- tive medicine after the 2018 intervention in any of the countries/regions. We observed a substantial number of concurrent pregnancies during valproate exposure, but with a declining rate after the 2018 RMMs in Tuscany, Italy (0.70 per 1000 valproate users pre- and 0.27 post-intervention), Spain (0.48 and 0.13), the Netherlands (0.34 and 0.00), and an increasing rate in UK (1.13 and 5.07). Conclusion There was a small impact of the 2018 RMMs on valproate use in the studied European countries/regions. The substantial number of concurrent pregnancies with valproate exposure warrants a careful monitoring of implementation of the existing PPP for valproate in clinical practice in Europe, to see if there is any need for additional measures in the future. * Olaf Klungel o.h.klungel@uu.nl Extended author information available on the last page of the article Vol.:(0123456789) 690 S. Abtahi et al. prescribers and patients on the risks associated with val- Key Points proates during pregnancy. A review by the EMA’s Pharmacovigilance Risk Assess- Valproate use has generally declined among females of ment Committee (PRAC) in 2018 showed that despite the childbearing potential in Europe between 2010 and 2020. 2014 RMMs many patients were still not informed on the After the 2018 EU risk minimisation measures for risks of using valproates during pregnancy, a high level valproate, a significant decline in valproate utilisation of exposure to sodium divalproate and valpromide among was observed among females of childbearing potential in females of childbearing potential persisted and prescribing Tuscany (Italy), Spain, and UK, while there was a non- conditions were not adhered to, especially in the bipolar significant decline in the Netherlands, compared to the disorder indication [7]. In March 2018, the PRAC rec- time period before. ommended an update of the RMMs including the intro- duction of a pregnancy prevention programme (PPP) [8]. The contraceptive coverage during valproate use was This included an assessment of each patient’s potential to low across countries/regions, and increased only in the become pregnant, pregnancy tests before starting and dur- Netherlands after the 2018 intervention. ing treatment with valproates, counselling about the risks Switching from valproate to alternative medicine (for of valproate treatment, the need for effective contraception indications of valproate) was not increased after the throughout treatment, an annual review of ongoing treat- 2018 risk minimisation measures in any of the studied ment by a specialist and changes to the product information countries/regions. including a visual boxed warning. Furthermore, according to the 2018 measures by PRAC, valproates are contraindicated There was a substantial number of concurrent pregnan- in females of childbearing potential for all indications (epi- cies with valproate exposure in the studied countries/ lepsy, bipolar disorders and prophylaxis of migraine attacks) regions. unless the conditions of a PPP, which has to be implemented in all EU Member States, are fulfilled. The aim of this study was to investigate the use of val- proates authorised in the EU before and after implementa- tion of the 2018 revised measures for pregnancy prevention in clinical practice, and effectiveness of the 2018 RMMs, 1 Introduction using longitudinal data collected in five electronic health- care databases (EHDs) from five European countries. To Valproate and related substances are licensed for the treat- answer this, we completed various objectives to measure the ment of epilepsy and acute mania in patients with bipolar utilisation of valproates, contraceptive coverage, alternative disorder in Europe, and also as prophylaxis for bipolar dis- medication use and occurrence of concurrent pregnancies orders and migraine headaches in some EU member states. with valproate exposure in females of childbearing potential. The teratogenic risk associated with the use of valproates in pregnant women are well established [1, 2]. Previous stud- ies showed that > 10% of children exposed to valproates in 2 Methods utero had a congenital malformation, and between 30 and 40% showed degrees of neurodevelopmental or behavioural 2.1 Setting and Data Sources disorders at an older age [3, 4]. In comparison with other antiepileptics among women with similar indications, chil- This was a retrospective cross-sectional time series study, dren exposed in utero to valproate had a two- to six-fold using data from five data sources in five European coun- increased risk of congenital malformations [2]. The risk is tries. A detailed description of the included data sources is dose dependent with no safe threshold below which no risk provided in Online Resource S.1.a. In short, this included exists [5]. Danish National Registries (DNR) that consists of adminis- In 2014, the European Medicines Agency (EMA) trative and clinical registers in Denmark with information on announced recommendations and set in place risk mini- all hospitalisation episodes, dispensations, birth records and misation measures (RMMs) to warn against valproate use vital statistics with a nationwide coverage of about 5.8 mil- in all females of childbearing potential, especially during lion individuals [9–11] (please see Online Resource S.1.b). pregnancy [6]. The RMMs included a set of educational The Agenzia regionale di sanità della Toscana (ARS Tus- materials and Direct Healthcare Professional Communica- cany) is a regional administrative claims database in Italy, tion (DHPC) to raise the understanding and awareness of with information on outpatient prescriptions, hospital admis- sions, diagnostic tests and exemptions from co-payment, Impact of 2018 EU Risk Minimisation Measures on Valproate Utilisation 691 linked to a birth registry, with a coverage of 3.6 million implemented by combining episodes when the gap between inhabitants. The PHARMO Data Network is an electronic them was shorter than 30 days. medical record (EMR) database with data linked from differ- Based on the constructed treatment episodes, subjects ent primary and secondary health care settings in the Neth- were counted as prevalent (or current) users of valproate erlands (such as general practitioner [GP] visits, in- and out- in a month if they had a treatment episode that overlapped patient pharmacies, hospitals, and perinatal registry), with with that month by at least one day. A person was counted as a coverage of 4.2 million active patients [12, 13]. The Base an incident (or new) user in a specific month if they started de Datos para la Investigación Farmacoepidemiológica en a first ever treatment episode that month since the start of Atencion Primaria (BIFAP) is an EMR database from mul- study in 01.2010, also considering the 1-year look-back tiple regions in Spain, with information on GP visits, pre- period of 2009. Discontinuation of valproates was defined as scriptions/dispensations, diagnostic procedures, laboratory no record of prescription/dispensing within 90 days follow- tests and specialist referrals, with a coverage of 9.4 million ing the theoretical end of the last valproate prescription/dis- active patients [14]. The Clinical Practice Research Data- pensing within a valproate episode. Females meeting criteria link (CPRD) is a primary care database from the UK, with for discontinuation may re-initiate, leading to multiple epi- data on all GP visits, prescriptions and laboratory tests from sodes of treatment. Duration of use was defined as the time ~10 million active patients enrolled [15]. All these covers from initiation of treatment based upon the first recorded a source population of > 30 million persons with approxi- prescription/dispensing in the look-back or study periods mately 15–20 million females of childbearing potential. until a specific month. Treatment duration per episode was stratified as follows: < 6 months, 6–12 months, and > 1 year. 2.2 Study Population 2.3.2 Alternative Medications All females of childbearing potential (aged 12–55 years) Alternative medications used for epilepsy were includ- between 01.01.2010 and 31.12.2020 were included from five ing: Carbamazepine, phenobarbital, phenytoin, primidone, databases (up to 31.12.2018 for DNR). The definitions of the clobazam, clonazepam, eslicarbazepine acetate, lamotrigine, study population, exclusion criteria applied, and main study oxcarbazepine, perampanel, rufinamide, topiramate, zon- time points are shown in Online Resource S.1.c. Women isamide, brivaracetam, ethosuximide, gabapentin, lacosa- entered the cohort on the latest of 01.01.2010 (having one mide, levetiracetam, pregabalin, tiagabine, vigabatrin. Alter- year of previous valid data), the 12th birthday and database natives for bipolar disorder treatment (as maintenance) were: registration. The cohort exit was defined as the earliest of Lithium, quetiapine, olanzapine or lamotrigine. Alternatives 31.12.2020 (study end), the 56th birthday, sterilisation, for migraine prophylaxis were: beta-blockers, topiramate, death and database deregistration. amitriptyline, flunarizine, pizotifen, clonidine. Prescrip- tion or dispensing events of alternative medications were 2.3 Exposures of Interest extracted from drug files by ATC/CPRD product code or product name. 2.3.1 Valproates The occurrence of switches from valproate to alternative medications was defined as the occurrence of a prescription Valproate-containing medicinal products (valproic acid, of an alternative medication during an episode of valproate sodium valproate, valproate pivoxil, valproate semi-sodium, use or within the discontinuation period. If an alternative valpromide, valproate bismuth, calcium valproate, valproate medication started after valproate discontinuation (after 90 magnesium) were the main study exposure. Prescription or days of the episode end), this was not considered as a switch. dispensing events were extracted from drug files by ATC/ CPRD product code or product name. Treatment episodes 2.4 Outcomes of Interest were separately constructed for valproate exposure fol- lowing an existing methodology [16, 17]. An episode for 2.4.1 Contraception a product with a given ATC/BNF code started on the date of incident prescription/dispensing. Each data access part- Contraception was defined as at least one user-independent ner (DAP) recommended the approach that was expected method applied by the woman, or a hormone-based method to minimise exposure misclassification in their database, combined with a barrier method. As the barrier method can- given their available data (Online Resource S.1.d). Overlap not be assessed reliably, it was not considered. Contraception between prescription refills of valproates was accounted for coverage episodes were constructed based upon ATC/BNF by adding the overlapping days to the end of the treatment and procedure codes, prescription dates, units prescribed for episode. A permissible gap of 30 days between episodes was each method, and a fixed assumed duration for each method. 692 S. Abtahi et al. Hormone-based user-dependent methods consisted of: vagi- dates for the implementation of the PPP were, respectively, nal ring (28 days, including 21 days using and one week 16.07.2018 and 11.10.2018 in Denmark, 08.08.2018 and off), contraceptive patch (28 days, using for 3 weeks, one 02.10.2018 in Italy, 10.08.2018 and 12.12.2018 in the week off), progestogen-only pill or desogestrel progestogen- Netherlands, 24.07.2018 and 01.12.2018 in Spain, and only pill (28 days continuously), and combination pills (28 30.04.2018 and 31.07.2018 in the UK. days, as 21 days using and one week off). User-independent non-permanent methods included: contraceptive implant (3 years), contraceptive injection (12 weeks), intrauterine 2.6 Strategy for Distributed Analyses device (3 years), intrauterine system (3 years). User-inde- pendent permanent methods were female sterilisation and This study has been conducted in a distributed manner using hysterectomy. Because of no family linkage available from the ConcePTION Common Data Model (CDM) and com- data sources, male sterilisation (vasectomy) was not consid- mon analytics [20, 21]. This approach was taken to support ered. Subjects were censored when a record of sterilisation the standardisation of analyses across data sources in the or hysterectomy was observed, as they no longer were in the study and to improve transparency of the process of evi- population of females of childbearing potential. dence generation. According to the ConcePTION CDM pipeline, codes are linked to concepts but remain in their 2.4.2 Pregnancies original format. Then, an Extract, Transform and Load (ETL) process converts original data to the ConcePTION Pregnancies were identified across databases using a preg- CDM tables, assisted by an ETL standard template defin- nancy algorithm developed by Gini et al within the frame- ing the link between source data and the target tables of the work of the ConcePTION project (https://​www.​imi-​conce​ CDM [22]. All intermediate data sets remained local with ption.​eu/). This builds directly on top of a published algo- each DAP, and data processing and analysis steps were per- rithm for detecting pregnancies [18]. Briefly, the proposed formed by DAPs using common analytic scripts provided pregnancy algorithm allowed the identification of past and by the core team. Only the results of the data quality checks ongoing pregnancies from four main streams: perinatal or and analyses were uploaded to a central, secure platform of birth registries, administrative data banks using diagnosis Utrecht University (YODA). All analyses scripts are open codes, and a tailored-combined stream, which uses addi- source and are publicly available at: https://g​ ithub.c​ om/L ​ ot4/​ tional data from medical observations (item-sets). Lot4S​tudies. The algorithm first identifies pregnancies from any pos- sible records and subsequently establishes the start and end date of pregnancy by processing all the available informa- 2.7 Statistical Analyses tion on a hierarchical manner. Hierarchy is based on the quality of records with four colour indicators: “Green” was Descriptive statistics were provided on the source and study assigned when both pregnancy start and end dates were population in addition to baseline characteristics of valproate recorded in the database; “Yellow” shows that pregnancy users. Monthly prevalent and incident use of valproates were end date was recorded but pregnancy start date was imputed; estimated over the study period, for each DAP, and strati- “Blue” indicates that pregnancy start date was recorded fied by age group, indication and duration of treatment. The in the database and end date was imputed; and the colour change in overall trend in current (prevalent) use of valproate code “Red” represents a pregnancy event for which both was calculated across the study period for all centres. Fur- pregnancy start and end dates were imputed. Hence, green thermore, the numbers of discontinuers of valproates were shows the highest quality and red indicates the lowest one counted per month, for each DAP, and stratified by age to ascertain a pregnancy event from all data available in the group, indication, and duration of treatment. Compliance database. This framework was applied for ARS, PHARMO of prescribers with 2018 RMMs for valproate was analysed and BIFAP while CPRD used a different strategy [19]. How- with the monthly proportion of valproate use during an ever, PHARMO was only able to include in the algorithm episode of contraception, per each DAP, and stratified by the perinatal registry except in the first few years of the study age group, indication and duration of valproate use. Use of period (Online Resource S.2). alternative medications in prevalent users of valproate was counted during the study period as the monthly number of 2.5 PRAC Intervention prescriptions/dispensings, for each DAP, and reported sepa- rately for each of the three indications for valproates. The Although EMA released a statement on the recommenda- monthly incidence of treatment switches was estimated as tions and PPP in March 2018, the actual implementation the number of women who switched in a month divided by dates varied across countries. The estimated start and end the total number of current valproate users that month. Impact of 2018 EU Risk Minimisation Measures on Valproate Utilisation 693 To test the effect of the 2018 RMMs on outcomes, inter- 2.8 Data Management and Quality Checks rupted time series (ITS) analyses were conducted. Seg- mented generalised least squares regression analysis was The study was conducted according to the guidelines for used to compare the pre-intervention (2010–2018) and Good Pharmacoepidemiology Practice [24], and accord- post-intervention (2018–2020) level and trend changes in ing to the ENCePP code of conduct [25]. It bears an each of the outcome measures mentioned above. The cut- ENCePP Seal and is registered in the EU PAS register off point for intervention to run the ITS analysis was set as (EUPAS31001). the first month of the implementation window (mentioned We conducted three levels of data quality and charac- in 2.5), and the remaining months of the implementation terisation checks to ensure that the ETL process has been window were excluded from the ITS model to avoid min- properly conducted and the analytic datasets produced by gling of the pre- and post-intervention periods. Based on DAPs are complete and valid enough to be used in the main the actual implementation windows and for consistency, analyses. Level 1 data checks reviewed the completeness a 2- or 3-month intervention period was chosen for ITS and content of each variable in each table to ensure that the models, as Aug–Sep 2018 for ARS Tuscany, Aug–Oct required variables contain data and conform to the formats 2018 for PHARMO, Aug–Oct 2018 for BIFAP, and specified by the CDM specifications (e.g., data types, vari- May–Jul 2018 for CPRD. A slope and level-change impact able lengths, formats, acceptable values, etc.). Level 2 data model with two segments was used: segment 1 modelled checks assessed the logical relationship and integrity of data constant pre-intervention outcome, segment 2 modelled values within a variable or between two or more variables, post-intervention outcome, by estimating the regression within and between tables. And Level 3 data checks exam- coefficients and p values. The beta coefficients of the ITS ined data distributions and trends over time, by examining analyses produced for each DAP were presented using for- output by year and month, for each DAP’s databases. To est plots. Additionally, the impact of the 2018 RMMs was help with evaluating the output of these checks, especially quantified for various outcome measures using the beta tailored quality check sheets were used in weekly 1–1 meet- coefficients of ITS analyses, in form of mean rate differ- ings between DAPs and the core team. ence post-intervention, p value and percentage of change compared to the mean counterfactual value had the inter- vention not occurred [23]. 3 Results The occurrence of a pregnancy event with a valproate exposure have been assessed in two ways: in “Analytic A total of 9,699,371 females of childbearing potential from method A”, we counted the monthly occurrence of a val- the five participating centres were included in the study proate prescription/dispensing during a pregnancy time population between 2010 and 2020 (until 2018 for DNR, window, as medication use during pregnancy. Here we and 2019 for PHARMO), where 69,533 individuals had used counted only the first date of a valproate prescription/ valproates at least once during the study period. The largest dispensing and removed any other prescription/dispens- subpopulation of valproate users was from ARS Tuscany ing dates to avoid duplicate counting of events during (N = 29,093), being also the highest percentage of its total the same pregnancy for each unique patient. In “Analytic population (2.6%) (Table 1). The median follow-up time method B”, we counted the monthly occurrence of start of ranged between 4.4 years (interquartile range [IQR] 3.9 a pregnancy during a treatment episode of valproate. For years) for CPRD and 11.0 years (IQR 4.4 years) for ARS this we did not remove multiple potential pregnancies that Tuscany. The mean age at the start of follow-up was always might have occurred in the same patient during a (long) > 30 years, ranging between 33.9 years (for DNR) and 37.0 treatment with valproate, as each could represent the tera- years (for CPRD). Similar across all databases, most patients togenic risk to a unique foetus. In both methods, rates were were started to follow up at the highest age stratum (i.e., produced per number of prevalent users of valproates each 41.0–55.0 years). month, and also aggregated counts and rates per each year The utilisation of valproates generally declined across and for the whole study period before and after the 2018 the study period and in most centres. While there was a RMMs, for each database. 2.2% increase in prevalent use of valproates in ARS Tus- Additionally, sensitivity analyses were conducted by cany (between 01.2010 and 12.2020), a 24.4%, 29.1%, and setting the start of pre-intervention time period for ITS 37.7% decline was observed in current use, respectively, in analyses to 01.01.2015 (instead of 01.01.2010), to mini- PHARMO (01.2010 and 12.2019), BIFAP (01.2010 and mise any effect from the 2014 RMMs on the overall trends, 12.2020), and CPRD (01.2010 and 10.2020). The monthly and restricting the end of the study period to 02.2020, to prevalent use of valproate ranged between 1.9 and 2.2 per consider the impact of the COVID-19 pandemic on the 1000 women of childbearing potential in DNR, between 6.1 time period afterwards. and 7.7 in ARS database, between 1.2 and 1.6 in PHARMO 694 S. Abtahi et al. Table 1  Baseline characteristics of the valproate users within the different databases in Denmark, Italy, the Netherlands, Spain, and the UK Denmark, DNR Italy, ARS Tuscany The Netherlands, Spain, BIFAP UK, CPRD PHARMO Total number of study population 1,575,216† 1,117,251 591,500 5,066,393 1,349,011 Number of valproate users (% of total 9159 (0.6%) 29,093 (2.6%) 2725 (0.5%) 22,325 (0.4%) 6231 (0.5%) population) Median follow-up, years (IQR) 8.7 (2.0) 11.0 (4.4) 10.0 (1.4) 10.0 (3.9) 4.4 (3.9) Age Mean age at index date* (years, SD) 33.9 (12.4) 36.5 (11.3) 34.4 (12.3) 34.7 (12.1) 37.0 (11.7) 12.0–20.9 years (%) 1801 (19.7%) 3648 (12.5%) 510 (18.7%) 3630 (16.3%) 699 (11.2%) 21.0–30.9 years (%) 1749 (19.1%) 4216 (14.5%) 455 (16.7%) 3903 (17.5%) 1098 (17.6%) 31.0–40.9 years (%) 2281 (24.9%) 8612 (29.6%) 693 (25.4%) 6362 (28.5%) 1554 (24.9%) 41.0–55.0 years (%) 3328 (36.3%) 12,617 (43.4%) 1067 (39.2%) 8430 (37.8%) 2880 (46.2%) CPRD Clinical Practice Research Datalink, DNR Danish National Registries, IQR interquartile range, SD standard deviation *Index date is defined as the date of entry into the study (the latest of 01 January 2010, the date the individual entered the underlying population of data source, or the individual's ­12th birthday) † Baseline numbers in Denmark are calculated from a different source (please see Online Resource S.1.b). The cumulative number here shows the included females of childbearing age in 01.2010 database, between 1.3 and 2.0 in BIFAP, and between 1.9 summarise the beta coefficients of ITS analyses of both main and 3.2 in CPRD. Additionally, we observed a declining and sensitivity analyses for a level and trend change in preva- trend in incident use of valproate in all databases across the lent use after the 2018 RMMs across centres. study period (2010–2020). There was no statistically significant change in level or The stratified analyses by age showed that most of the trend of incident use seen in any of the studied databases prevalent users of valproates were in the age stratum 41–55 after the implementation of the 2018 EU RMMs in the years (Online Resource S.3.a). The stratification by treat- main models (Online Resource S.3.b). Quantification of the ment duration showed that there was an observable trend for impact showed a 6.0%, 12.7%, 39.3% increase in valproate a lower proportion of short-term users (≤ 6 months) and a incident use post-intervention versus before, respectively, in higher proportion of long-term users (> 1 year) towards the ARS Tuscany, PHARMO, BIFAP, and a 16.5% decrease in end study period in most databases. CPRD, although none were statistically significant (Table 2). There was a statistically significant change in the trend The only significant change in the sensitivity analyses was of prevalent use of valproate after the implementation of observed in PHARMO for an increase in level immediately the 2018 RMMs in most of the studied regions/countries, after the intervention. including ARS Tuscany, BIFAP and CPRD, without a sig- The monthly proportion of discontinuers ranged between nificant change in the level (Fig. 1b–e). The quantification 1.3 and 2.6% in DNR, 4.4–7.5% in ARS, 1.5–5.3% in of the impact showed that the prevalent use of valproate PHARMO, 3.2–5.2% in BIFAP and 0.8–3.2% in CPRD. decreased in average by 7.7% after the 2018 intervention The ITS analyses showed that there was a significant change versus before in ARS Tuscany (mean difference post-inter- in both level and trend of valproate discontinuers in ARS vention −0.590 per 1000 person months, p value < 0.01), Tuscany (0.6% mean difference post-intervention, p value while this decline was 11.3% in BIFAP (− 0.198, < 0.01), = 0.81) and CPRD (− 28.8%, < 0.01), but no significant and 5.9% in CPRD (0.143, 0.02) (Table 2). In PHARMO, changes in PHARMO (9.1%, 0.29) and BIFAP (6.2%, 0.31) we observed a non-significant decline of 3.1% in prevalent after the 2018 EU RMMs versus before (Table 2, and Online use post-intervention (− 0.043, 0.19). It was not possible Resource S.3.c). Running the sensitivity analysis yielded to model the ITS analyses on data from DNR, because not non-significant estimates for a change in proportion of val- enough data points were available here after the 2018 RMMs proate discontinuers after the 2018 RMMs in ARS Tuscany (Fig. 1a). Running the sensitivity analysis by excluding and PHARMO, but a significant change in level in BIFAP, the first five years and COVID-19 period (with follow-up and a significant change in trend in CPRD. between 01.2015 and 02.2020) resulted in similar findings Only between 5.8 and 9.3% of monthly valproate dispens- in ARS Tuscany, and PHARMO compared with the main ings in DNR, 11.9–22.5% in PHARMO, 0.2–3.3% in BIFAP, analyses (Fig. 1b, c), but a significant change in both level and 10.1–24.9% of valproate prescriptions in CPRD have and trend post-intervention in BIFAP (Fig. 1d) and non-sig- occurred during an episode of contraceptive use. The ITS nificant changes in CPRD (Fig. 1e). The forest plots in Fig. 2 analyses showed no significant change in level or trend of Impact of 2018 EU Risk Minimisation Measures on Valproate Utilisation 695 Fig. 1  Change in level and trend in rates of prevalent use of valproate in five participating centres over the study period, modelled by inter- rupted time series (ITS) analysis. The time of implementation of the 2018 risk minimisation measures for valproate is shown with a light grey bar (starts with a dashed black line), and the counterfactual trend (had the intervention not occurred) is shown with the dashed red line thereafter. Note that the end of study period differs per data source. In case of DNR, the trends of prevalent use over time have been shown instead, as an ITS modelling was not possible due to limited time points after the 2018 intervention proportion of compliant valproate dispensings after the 2018 intervention in the studied databases, in either the main or in the sensitivity analyses (Online Resource S.3.d). The quan- tification of the impact showed that there was a significant Fig. 1  (continued) 12.0% increase in compliant valproate dispensings with con- traceptive coverage after the 2018 RMMs in PHARMO (p = 0.02), a non-significant 6.5% decrease in BIFAP (0.39), and information on contraceptive use during (parts of) the study a significant 26.6% decrease in CPRD (< 0.01), compared period from ARS Tuscany, thus no trend has been shown. with the time before. In case of DNR, only the general trend The trend in alternative medication use per various indi- during the study period is shown. We could retrieve limited cations of valproates across databases are given in Online Resource S.3.e. In summary, the trends for epilepsy and 696 S. Abtahi et al. e. CPRD, UK PHARMO (Table 3). With “Analytic method A”, we found 386 first dates of a valproate dispensings in ARS Tuscany Current valproate users per 1000 persons that occurred during a pregnancy time window (with the quality colour codes green and yellow) before implementa- 5 4 tion of the 2018 RMMs, with a rate of 0.70 per 1000 val- 3 proate users. The count and rate of such concurrent events 2 1 in ARS declined to 40 and 0.27, respectively, after the 2018 0 2010-01 2011-07 2013-01 2014-07 2016-01 2017-07 2019-01 2020-07 RMMs. In PHARMO, we observed only 27 first dates of Date a concurrent valproate dispensing with a pregnancy event Level change: -0.021 current users/1000 persons, p=0.762 (all with a colour code green) with a rate of 0.34 per 1000 Trend change: -0.008 current users/1000 persons, p=0.042* users, in the whole study period before 2018 RMMs, while there was no single event in the time period after. In BIFAP, there were 330 prescriptions/dispensings of valproate dur- Current valproate users per 1000 persons ing a pregnancy episode (all with a quality colour code yellow) with a rate of 0.48 per 1000 drug users before the 5 4 2018 RMMs. The number of concurrent drug use/pregnancy 3 events reduced to 20 events with a rate of 0.13 in the time 2 1 period after. In contrast to other databases, we observed an 0 2015-01 2015-10 2016-07 2017-04 2018-01 2018-10 2019-07 increasing rate of simultaneous valproate prescriptions dur- Date ing pregnancy in CPRD, where 204 such events happened Level change: 0.016 current users/1000 persons, p=0.778 before 2018 RMMs (with a rate of 1.13 per 1000 users) and 56 events post-2018 intervention (rate 5.07). The findings of Trend change: 0.004 current users/1000 persons, p=0.331 * ITSA not conducted due to limited post-intervention data available. the “Analytic method B” (i.e., start of a pregnancy during a Fig. 1  (continued) treatment episode of valproates) were similar to those men- tioned above (Table 3). There were no data on pregnancy counts available from DNR. bipolar disease drugs were generally increasing in DNR, The time distance between start of a pregnancy and pre- ARS, and PHARMO during the study period, while the scription/dispensing date of a valproate during a pregnancy trend in use of migraine drugs remained steady. The trends episode has been calculated in some DAPs. In ARS Tuscany, in use of alternative medications for all indications were the median distance between a pregnancy start and dispens- levelling in BIFAP, and increasing in CPRD over the study ing date of a valproate was 27 days (IQR 48 days, maximum period. 278 days). While in PHARMO, the median length was 38 The monthly rate of switching from valproates to alterna- days (IQR 63, maximum 272), and in CPRD the median was tive medications during the study period ranged between 0.9 19 days (IQR 23, maximum 277). and 3.2% in DNR, 1.9–4.5% in ARS Tuscany, 0.7–3.0% in PHARMO, 1.0–3.7% in BIFAP, and 0.9–3.2% in CPRD. The ITS analysis of switching rates from valproates to alternative 4 Discussion medicine before versus after the 2018 RMMs revealed that there was a statistically significant change in level and trend This study investigated the use of valproates before and after of switchers in ARS Tuscany, and non-significant changes the 2018 RMMs and the effectiveness of the revised PPP in in BIFAP (Online Resource S.3.f). The sensitivity analyses females of childbearing potential in five European countries. yielded non-significant changes in both centres. The quanti- The utilisation of valproates, both incident and prevalent use, fication of the impact showed that there was a 3.3% decline generally declined across the study period (2010–2020) in in switching rates from valproate to alternative medicine all centres, besides a small increase of prevalent use in ARS after the 2018 RMMs versus before in ARS (p = 0.47), and Tuscany (Italy). We observed a significant decline in preva- 14.8% decrease in BIFAP (0.22), both statistically non-sig- lent use of valproate in ARS Tuscany (Italy), BIFAP (Spain) nificant. It was not possible to model the ITS analysis in and CPRD (UK), and a non-significant decline in PHARMO PHARMO and CPRD because the counts of switchers after (Netherlands) after the 2018 RMMs compared to the period 2018 were too small to fit a stable model, and in DNR there before, but neither a significant decline in incidence rates, were insufficient time points after the 2018 intervention. nor a significant increase in valproate discontinuation. This In general, we observed a substantial number of concur- discrepancy might be explained by a general decline in the rent pregnancies with a valproate exposure in ARS Tus- prevalent use of valproate since 2010 due to prior (partly cany, BIFAP and CPRD, and fewer concurrent events in successful) RMMs and specialists being discouraged from Impact of 2018 EU Risk Minimisation Measures on Valproate Utilisation 697 Table 2  Quantification of the impact of the 2018 risk minimisation proates, discontinuation of valproates, contraceptive coverage during measures derived from the ITS analysis (of main models) on vari- valproate use, and switching from valproates to alternative medica- ous outcome measures, including incident and prevalent use of val- tions, across the participating centres ARS Tuscany, IT PHARMO, NL BIFAP, ES CPRD, UK Mean difference p value Mean difference p value Mean difference p value Mean difference p value post-intervention* post-intervention* post-intervention* post-intervention* (%) (%) (%) (%) Incident use of val- 0.009 (6.0) 0.53 0.002 (12.7) 0.48 0.003 (39.3) 0.37 − 0.002 (− 16.5) 0.43 proates (per 1000 person months) Prevalent use of val- −0.590 (−7.7) < 0.01 −0.043 (−3.1) 0.19 − 0.198 (− 11.3) <0.01 − 0.143 (− 5.9) 0.02 proates (per 1000 person months) Discontinuation of 0.032 (0.6) 0.81 0.244 (9.1) 0.29 0.255 (6.2) 0.31 −0.422 (− 28.8) <0.01 valproates (per- centage per month) Contraceptive NA 2.114 (12.0) 0.02 − 0.186 (− 6.5) 0.39 −6.196 (−26.6) <0.01 coverage during valproate use (per- centage per month) Switching from −0.092 (−3.3) 0.47 NA − 0.276 (− 14.8) 0.22 NA valproates to alter- native medications (percentage per month) Because no ITS model could be fit for DNR, due to limited time points after the 2018 intervention, such quantification of the impact was not possible here CPRD Clinical Practice Research Datalink, DNR Danish National Registries, ES Spain, IT Italy, ITS interrupted time series, NA not available, NL the Netherlands, UK the United Kingdom Statistically significant rate ratios are shown in bold *The rates of mean difference post-intervention show the change in various outcome measures after the 2018 risk minimisation measures, com- bining the level and trend change from the ITS model. The percentage in parenthesis shows the mean percent change of rates of various outcome measures after the 2018 intervention compared to the mean counterfactual values had the intervention not occurred initiating valproate in females of childbearing potential. A Finally, there was a substantial number of concur- changing pattern in treatment duration in most centres, with rent pregnancies with valproate exposure in ARS (Italy), lower proportion of short-term (≤ 6 months) and a higher BIFAP (Spain) and CPRD (UK), with fewer events in proportion of long-term valproate users (> 1 year) towards PHARMO (Netherlands). The occurrence of concur- the end of study period, can partly explain this hypothesis. rent pregnancies in PHARMO (Netherlands) may be Additionally, the rate of contraceptive coverage at the underestimated, as most of the study period pregnan- start of valproate treatment was low across all DAPs, as cies were identified only from a perinatal registry (i.e., always below a quarter of new valproate treatment episodes pregnancies ending in a later stage). On the other hand, had started during contraceptive use. The only significant the higher occurrence of valproate exposure during preg- increase in compliant valproate prescriptions/dispensings nancy in ARS (Italy), BIFAP (Spain) and CPRD (UK) with a contraceptive coverage after the 2018 RMMs was could mean that they captured pregnancies ending pre- observed in PHARMO (Netherlands). Regarding the alterna- maturely. The rate of concurrent pregnancies during val- tive medications, we found an increasing trend in rates of use proate use declined in ARS (Italy), BIFAP (Spain) and for the indications epilepsy and bipolar diseases across the PHARMO (Netherlands) after the 2018 intervention, but study period in most databases (i.e., DNR [Denmark], ARS not in CPRD. This may suggest that often assessing that a [Italy], PHARMO [Netherlands] and CPRD [UK]), while pregnancy has started in a woman exposed to valproate or the rates for migraine were mostly steady. The monthly rate assessing that a pregnant woman has been exposed to val- of switch from a valproate to an alternative medication was proate, results in termination of the pregnancy. The higher similar across all DAPs (< 5%), with no significant increase rate of concurrent pregnancies in CPRD (UK) before ver- in switching rates after the 2018 intervention versus the time sus after the 2018 RMMs may be due to the known attri- before. tion of the CPRD GOLD databases in recent years and 698 S. Abtahi et al. Fig. 2  Forest plots showing the level and trend beta coefficients of the interrupted time series analyses of prevalent use of valproate after the implementa- tion of the 2018 risk minimisa- tion measures across centres, in the: a main, and b sensitivity analysis artificial inflation of rates with smaller denominators (loss Germany, Spain, Sweden and UK) found limited effective- to follow-up) towards the end of study period. ness of the measures based on the varying proportions of A drug utilisation study that evaluated the 2014 RMMs valproate initiation as second-line therapy from pre- to for valproate use in five European countries (France, post-implementation period, and a decrease in incidence of Impact of 2018 EU Risk Minimisation Measures on Valproate Utilisation 699 pregnancies exposed to valproate only in Sweden and UK 2010–Jul. 2018) and post-2018 RMMs (i.e., Aug. 2018 – end of study period for each database) in each database, which represent the sum of unique patients who were exposed each month to The number of prevalent users of valproates, used as the denominator for the rates here, is actually the sum of monthly prevalent use counts for the time period pre-2018 RMMs (i.e., Jan. Table 3  Aggregated concurrent pregnancy events during valproate exposure, stratified by the analytic method (A: valproate use during a pregnancy time window, and B: pregnancy occurrence Noteworthy, only those pregnancy events with a quality colour code green and yellow have been considered in this table. Also, the observed concurrent events in the two analytic methods (A pregnancies of valproates† 1000 user pregnancies of ­valproates† 1000 user pregnancies of ­valproates† 1000 user pregnancies of ­valproates† 1000 user Concurrent Prevalent users Rate per Concurrent Prevalent users Rate per Concurrent Prevalent users Rate per Concurrent Prevalent users Rate per [26]. However, the value of these findings remains limited *The cut-off point for a demarcation between pre- and post-intervention time periods was selected as Aug. 2018 for all DAPs, based on the implementation period of RMMs across countries 1.13 5.07 1.10 6.16 due to the descriptive nature of the study and reliance on absolute counts instead of standardised rates. Running an ITS model on the UK data from the same group showed that there was no statistically significant change in period 11,043 181,054 11,043 181,054 (level) or trend in the proportion of valproate initiation as second-line therapy before versus after the 2014 RMMs [27]. Another study in France reported an overall decrease CPRD, UK in valproate use among females of childbearing potential with epilepsy after the 2014 RMMs, in an ITS study with a 204 56 200 68 control group (males with epilepsy) [28]. There are few studies that focused on the risk awareness and adherence with EU RMMs for valproates in 2014 and 0.48 0.13 0.45 0.13 2018. A cross-sectional survey among physicians across France, Germany, Spain, Sweden and UK concluded that a majority of physicians had good knowledge of prescribing valproates in females of childbearing potential, where 96% 157,981 694,549 157,981 694,549 of physicians responded that they only prescribe valproate to during a valproate treatment episode) and by time period of pre- and post-2018 risk minimisation measures, across databases women with epilepsy or bipolar disorders when other treat- ments were ineffective/not tolerated [29]. Another study on BIFAP, ES the risk awareness and adherence with the 2018 RMMs in eight European centres (in Belgium, Denmark, Greece, Lat- 330 20 316 21 via, Portugal, the Netherlands, Slovenia and Spain) found that there was high awareness of the teratogenic risks asso- ciated with valproate use during pregnancy among patients 0.34 0.30 ES Spain, IT Italy, NL the Netherlands, RMM risk minimisation measures, UK the United Kingdom (71%), prescribers (94%), and pharmacists (95%) [30]. How- 0 0 ever, in both these qualitative studies, the use of educational materials and information tools (such as patient card, risk acknowledgement forms, and healthcare professional guides) 10,896 79,197 10,896 79,197 was low, and according to respondents, the most beneficial PHARMO, NL tool was the warning symbol on packaging of drugs. These results can complement what we found in this quantitative study with declining trend in utilisation of valproates since 2010 and after the 2018 RMMs, but still low rates of cover- 27 0 24 0 age by contraceptive methods. It is of note that valproate RMMs have been in place for B. Start of pregnancy during a treatment episode of valproate 0.70 0.27 0.68 0.33 many years [7]. The referral in 2014 further strengthened the restrictions in use of valproate in females of childbearing potential, and this has become even more strengthened in A. Valproate use during a pregnancy time window 2018 [8]. A clear cut-off in 2018 to detecting an effect might 150,769 552,506 150,769 552,506 not be visible on all outcomes studied, as for example the advice to use contraceptive measures was already in place ARS Tuscany, IT before 2018, and the use of valproate in females of child- bearing potential had been declining since 2010. While our findings confirmed this latter finding (a declining trend in Pre-2018 RMMs* 386 Post-2018 RMMs* 40 Pre-2018 RMMs* 374 Post-2018 RMMs* 50 valproate utilisation between 2010 and 2020), the influence and B) are overlapping from previous RMMs may have prevented us from detecting a significant change for some of the outcomes studied. This study had several strengths. Selection bias was miti- gated by including all females of childbearing potential reg- valproate istered in each database at any time during the study period (with a total sample size of ~10 million), in addition to † 700 S. Abtahi et al. setting minimal exclusion criteria. Using several databases implementation of 2018 RMMs in Europe, an increase in from different countries across the continent has led to a compliant valproate prescriptions/dispensings with a con- large and diverse study population with good representation traceptive coverage only in PHARMO (Netherlands), no and generalisability to a European and global setting. A key significant increase in switching rates from valproates to strength of this study was use of the ConcePTION CDM alternative medications, and declining rates of concurrent and common analytics to minimise unwanted heterogeneity pregnancy events with valproate exposure in most countries/ among databases, and the rigorous quality checks that were regions. But the substantial number of concurrent pregnancy conducted. Furthermore, to evaluate the implementation and events with valproate exposure warrants a careful monitor- the 2018 RMMs on valproate use, we benefited from the ITS ing of implementation of the existing PPP for valproate in analysis (a quasi-experimental design), which is one of the clinical practice in Europe, to see if there is a need for addi- strongest methods for studying the impact of an intervention tional measures in the future. on drug utilisation and other outcomes when conducting an Supplementary Information The online version contains supplemen- RCT is not feasible [31–33]. We were able to extract and use tary material available at https://d​ oi.o​ rg/1​ 0.1​ 007/s​ 40264-0​ 23-0​ 1314-3. a large number of data points and observations, especially before the intervention, which helped us to reduce any effect Acknowledgements The study report of this project to the European from fluctuations, variability and outliers in medication use Medicines Agency has been already published on Zenodo, as the fol- lowing: https://​doi.​org/​10.​5281/​zenodo.​70745​88 across the study period. Our study was not free from limitations. No single Euro- Declarations pean data source contains all the information required in this study, as the primary aim of data collection was not medical Funding The research leading to these results was conducted as part of research. By using retrospectively collected data from EHDs, the activities of the EU PE&PV (Pharmacoepidemiology and Pharma- covigilance) Research Network, which is a public academic partnership misclassification of exposure might have happened [34]. coordinated by the Utrecht University, the Netherlands. The project We had only dispensing information available from ARS, has received support from the European Medicines Agency under the BIFAP and PHARMO and GP prescriptions available from Framework service contract nr EMA/2018/28/PE. The content of this CPRD, which are 1 or 2 steps behind the actual drug use by paper expresses the opinion of the authors and may not be understood or quoted as being made on behalf of or reflecting the position of the patients, respectively, with no information on adherence to European Medicines Agency or one of its committees or working par- medication. Misclassification of outcome, especially in case ties. of pregnancy records, was another limitation. To minimise any issues with this, we used an adapted version of a previ- Conflicts of interest All authors declared the funding statement men- ously validated algorithm [12]. Also, lack of data on over- tioned above. Otherwise, SA, RP, JRA, MG, EA, VH, CB, SBK, PGP, the-counter contraceptives, which are not captured in any of HG, RG, AG, CH, LI, GL, MMP, OP, GR, PS, KW and OK declared none. CED is a board member of the International Society for Drug the databases, might have resulted in an underestimation of Bulletins, with no payments involved. MA declares grants or contracts the proportion of compliant valproate use with a contracep- from Novo Nordisk Foundation (NNF15SA0018404) to his institu- tive coverage. Moreover, not considering a longer look-back tion, and fees from Atrium, the Danish Pharmaceutical Industry As- period (> 1 year) might have limited our ability to ascertain sociation for leading and teaching pharmacoepidemiology courses. JB declares unrelated consultancy fees from CorEvitas LLC and WHO a true new (incident) use of valproate among the included Regional Office for Europe, and scholarship to attend ICPE 2022 from individuals, especially in the first years of study period. The International Society for Pharmacoepidemiology. CEH declares grants occurrence of the COVID-19 pandemic was a major obsta- or contracts from Novo Nordisk A/S, H. Lundbeck A/S, Ferring Phar- cle, where it shortened our post-intervention period. It also maceuticals and Leo Pharma, also funding of PhD projects to Depart- ment of Pharmacy, University of Copenhagen. EH and KMAS are em- brought us practical difficulties in accessing data from vari- ployees of the PHARMO Institute for Drug Outcomes Research. This ous databases, due to concurrent COVID-19 projects and independent research institute performs financially supported studies slower approval and releasing of data by competent authori- for government and related healthcare authorities and several phar- ties, particularly in case of DNR (Online Resource S.1.b). To maceutical companies. MS declares grants or contracts from Pfizer, AstraZeneca and Janssen to her institution, not related to this work. overcome this, we defined sensitivity analyses that excluded the COVID-19 period (after 02.2020). Ethics approval In Denmark, the Danish National Prescription Registry was accessed through the Research Service Unit of Statistics Denmark (FSEID-00004357/DST-project no. 707524), and approval for process- ing of personal health data was obtained through the UCPH (ref. no.: 5 Conclusions 514-0301/19-3000). In Italy, the participation to the EMA Lot4 study on valproates was included in ARS Toscana workplan and it did not In conclusion, there was a small impact of the 2018 RMMs require any further ethical approval. In the Netherlands, the EMA Lot4 on valproate utilisation and prescribing in the studied Euro- study on valproates has been approved for compliance with the general data protection regulation by the Institutional Review Board of ‘Sticht- pean countries/regions, considering a declining trend in ing Informatie voorziening voor Zorg en Onderzoek’ (dd. 10 Oct 2019). utilisation patterns of valproates since 2010 and after the Impact of 2018 EU Risk Minimisation Measures on Valproate Utilisation 701 In Spain, the favourable approval was received from BIFAP scientific 4. Reynolds EH, Green R. Valproate and folate: congenital and committee (reference num. 19_2019) on 23.02.2020, and from the developmental risks. Epilepsy Behav. 2020;108: 107068. Comité de Ética de la Investigación Con Medicamentos Regional de 5. Tomson T, Battino D, Bonizzoni E, et al. Comparative risk of la Comunidad de Madrid-CEIm-R (Acta 02/20) on 10.02.2020. In the major congenital malformations with eight different antiepileptic UK, the study proposal was reviewed and approved by the Independent drugs: a prospective cohort study of the EURAP registry. Lancet Scientific Advisory Committee of the Clinical Practice Research Data- Neurol. 2018;17(6):530–8. link (reference 20_009), which is responsible for reviewing protocols 6. European Medicines Agency. Valproate and related substances for scientific quality. Article-31 referral assessment report. 2014. http://​www.​ema.​ europa.​eu/​docs/​en_​GB/​docum​ent_​libra​r y/​Refer ​rals_​docum​ent/​ Consent to participate Not applicable. Valpr​oate_​and_​relat​ed_​subst​ances_​31/​Recom​menda​tion_​provi​ ded_ ​ by_ ​ P harm ​ a covi ​ g ilan ​ c e_ ​ R isk_ ​ A sses ​s ment_ ​ C ommi ​ t tee/​ Consent for publication Not applicable. WC500​177352.​pdf. Accessed 17 Nov 2022. 7. European Medicines Agency. Valproate and related substances. Availability of data and materials This multicentre study used https://​www.​ema.​europa.​eu/​en/​medic​ines/​human/​refer ​rals/​valpr​ anonymised patient data from electronic health care databases in dif- oate-​relat​ed-​subst​ances-0. Accessed 17 Nov 2022. ferent countries/regions. These data remained local within each centre 8. European Medicines Agency. New measures to avoid valproate and will not be distributed publicly. No additional data available. exposure in pregnancy endorsed. 2018. https://​www.​ema.​europa.​ eu/e​ n/d​ ocume​ nts/r​ eferr​ al/v​ alpro​ ate-a​ rticl​ e-3​ 1-r​ eferr​ al-n​ ew-m ​ easu​ Code availability All analyses scripts are open source and are publicly res-​avoid-​valpr​oate-​expos​ure-​pregn​ancy-​endor​sed_​en-0.​pdf. available at: https://​github.​com/​Lot4/​Lot4S​tudies. Accessed 17 Nov 2022. 9. Pedersen CB. The Danish civil registration system. Scand J Public Author contributions All authors contributed to intellectual concept Health. 2011;39(7 Suppl):22–5. and design of the research. SA, RP, CED, MG and EA wrote and 10. Schmidt M, Schmidt SA, Sandegaard JL, et al. The Danish checked the common analysis scripts for distributed analyses and con- National Patient Registry: a review of content, data quality, and ducted the final aggregated analyses. MA, SBK and CEH conducted research potential. Clin Epidemiol. 2015;17(7):449–90. the data acquisition and data analysis for Danish National Registries, 11. Kildemoes HW, Sørensen HT, Hallas J. The Danish National Denmark. RG, AG, CB, GL, OP and GR performed the data acquisi- Prescription Registry. Scand J Public Health. 2011;39(7 tion and data analysis for ARS Tuscany, Italy. EH and KS carried out Suppl):38–41. the data acquisition and data analysis for PHARMO, the Netherlands. 12. Bakker M, Herings R. PHARMO database network. In: Sturken- PGP, CH, MMP carried out the data acquisition and data analysis for boom M, Schink T, editors. Databases for pharmacoepidemiologi- BIFAP, Spain. JB and KW conducted the data acquisition and data cal research. Cham: Springer International Publishing AG; 2021. analysis for CPRD, UK. All authors contributed to data interpretation. p. 109–18. SA, RP, CED, JRA and VH were in charge of quality control processes. 13. Herings RM, Pedersen L. Pharmacy-based medical record linkage SA and RP wrote the manuscript. All authors were involved in editing systems. In: Strom B, Kimmel S, Hennessy S, editors. Pharma- the manuscript. All authors are accountable for their own contribution, coepidemiology. 5th ed. Hoboken: Wiley; 2012. p. 270–86. in addition to all aspects of the manuscript. All authors have approved 14. Maciá-Martínez MA, Gil M, Huerta C, et al. Base de Datos para the (re)submitted version of the manuscript. la Investigación Farmacoepidemiológica en Atención Primaria (BIFAP): a data resource for pharmacoepidemiology in Spain. Pharmacoepidemiol Drug Saf. 2020;29(10):1236–45. Open Access This article is licensed under a Creative Commons Attri- 15. Herrett E, Gallagher AM, Bhaskaran K, et al. Data resource pro- bution-NonCommercial 4.0 International License, which permits any file: clinical practice research datalink (CPRD). Int J Epidemiol. non-commercial use, sharing, adaptation, distribution and reproduction 2015;44(3):827–36. in any medium or format, as long as you give appropriate credit to the 16. Gardarsdottir H, Souverein PC, Egberts TC, et al. Construction of original author(s) and the source, provide a link to the Creative Com- drug treatment episodes from drug-dispensing histories is influ- mons licence, and indicate if changes were made. The images or other enced by the gap length. J Clin Epidemiol. 2010;63(4):422–7. third party material in this article are included in the article's Creative 17. Dima AL, Dediu D. Computation of adherence to medication and Commons licence, unless indicated otherwise in a credit line to the visualization of medication histories in R with AdhereR: towards material. If material is not included in the article's Creative Commons transparent and reproducible use of electronic healthcare data. licence and your intended use is not permitted by statutory regula- PLoS ONE. 2017;12(4): e0174426. tion or exceeds the permitted use, you will need to obtain permission 18. Matcho A, Ryan P, Fife D, et al. Inferring pregnancy episodes directly from the copyright holder. To view a copy of this licence, visit and outcomes within a network of observational databases. PLoS http://​creat​iveco​mmons.​org/​licen​ses/​by-​nc/4.​0/. ONE. 2018;13(2): e0192033. 19. Campbell J, Bhaskaran K, Thomas S, et al. Investigating the optimal handling of uncertain pregnancy episodes in the CPRD GOLD Pregnancy Register: a methodological study using UK References primary care data. BMJ Open. 2022;12(2): e055773. 20. Thurin NH, Pajouheshnia R, Roberto G, et al. From inception to 1. Schachter SC, Garcia P, Dashe JF. Antiseizure medications: ConcePTION: genesis of a network to support better monitoring Mechanism of action, pharmacology, and adverse effects. In: and communication of medication safety during pregnancy and UpToDate, Post, TW (Ed), UpToDate, Waltham, MA, 2021. breastfeeding. Clin Pharmacol Ther. 2022;111(1):321–31. 2. Weston J, Bromley R, Jackson CF, et al. Monotherapy treatment 21. Trifirò G, Coloma PM, Rijnbeek PR, et al. Combining multiple of epilepsy in pregnancy: congenital malformation outcomes in healthcare databases for postmarketing drug and vaccine safety the child. Cochrane Database Syst Rev. 2016;11(11): CD010224. surveillance: why and how? J Intern Med. 2014;275(6):551–61. 3. Wieck A, Jones S. Dangers of valproate in pregnancy. BMJ. 22. Gini R, Schuemie M, Brown J, et al. Data extraction and man- 2018;18(361): k1609. agement in networks of observational health care databases 702 S. Abtahi et al. for scientific research: a comparison of EU-ADR, OMOP, 29. Toussi M, Isabelle B, Tcherny-Lessenot S, et al. Effectiveness of mini-sentinel and MATRICE strategies. EGEMS (Wash DC). risk minimisation measures for valproate: a cross-sectional sur- 2016;4(1):1189. vey among physicians in Europe. Pharmacoepidemiol Drug Saf. 23. Travis-Lumer Y, Goldberg Y, Levine SZ. Effect size quantifica- 2021;30(3):283–91. tion for interrupted time series analysis: implementation in R 30. Leonardo Alves T, Hegger I, Almasdottir AB, et al. Measuring and analysis for Covid-19 research. Emerg Themes Epidemiol. awareness and use of pregnancy prevention measures for valproate 2022;19(1):9. and retinoid containing products among patients, prescribers and 24. Public Policy Committee, International Society of Pharmacoepi- pharmacists across eight European countries. 2022 (Submitted). demiology. Guidelines for good pharmacoepidemiology practice 31. Jandoc R, Burden AM, Mamdani M, et al. Interrupted time (GPP). Pharmacoepidemiol Drug Saf. 2016;25(1):2–10. series analysis in drug utilization research is increasing: sys- 25. Gini R, Fournie X, Dolk H, et al. The ENCePP Code of Conduct: tematic review and recommendations. J Clin Epidemiol. a best practise for scientific independence and transparency in 2015;68(8):950–6. noninterventional postauthorisation studies. Pharmacoepidemiol 32. Grimshaw J, Campbell M, Eccles M, et al. Experimental and Drug Saf. 2019;28(4):422–33. quasi-experimental designs for evaluating guideline implementa- 26. Toussi M, Shlaen M, Coste F, et al. Effectiveness of risk minimi- tion strategies. Fam Pract. 2000;17(Suppl 1):S11–6. sation measures for valproate: a drug utilisation study in Europe. 33. Been JV, Szatkowski L, van Staa TP, et al. Smoke-free legisla- Pharmacoepidemiol Drug Saf. 2021;30(3):292–303. tion and the incidence of paediatric respiratory infections and 27. Ehlken B, Nishikawa C, Kaplan S, et al. Effectiveness of risk wheezing/asthma: interrupted time series analyses in the four UK minimization measures for valproate: a drug utilization study nations. Sci Rep. 2015;14(5):15246. based on implementation of a risk minimization programme 34. European Medicines Agency: The European Network of Centres in Europe, analysis of data from the UK. Curr Med Res Opin. for Pharmacoepidemiology and Pharmacovigilance (ENCePP). 2022;38(3):461–8. Guide on Methodological Standards in Pharmacoepidemiology 28. Degremont A, Polard E, Kerbrat S, et al. Impact of recommenda- (Revision 8). EMA/95098/2010. http://​www.​encepp.​eu/​stand​ tions on sodium valproate prescription among women with epi- ards_​and_​guida​nces/​metho​dolog​icalG​uide.​shtml. Accessed 17 lepsy: an interrupted time-series study. Epilepsy Behav. 2021;125: Nov 2022. 108449. Authors and Affiliations Shahab Abtahi1 · Romin Pajouheshnia1 · Carlos E. Durán2 · Judit Riera‑Arnau2,3 · Magdalena Gamba1 · Ema Alsina2 · Vjola Hoxhaj2 · Morten Andersen4 · Claudia Bartolini5 · Sarah Brøgger Kristiansen4 · Jeremy Brown6 · Christine Erikstrup Hallgreen7 · Patricia Garcia‑Poza8 · Helga Gardarsdottir1,9 · Rosa Gini5 · Anna Girardi5 · Emily Holthuis10 · Consuelo Huerta11 · Luisa Ibánez3 · Giorgio Limoncella5 · Mar Martín‑Pérez8 · Olga Paoletti5 · Giuseppe Roberto5 · Patrick Souverein1 · Karin M. A. Swart10 · Kevin Wing6 · Miriam Sturkenboom2 · Olaf Klungel1 1 7 Division of Pharmacoepidemiology and Clinical Copenhagen Centre for Regulatory Science, Department Pharmacology, Utrecht Institute for Pharmaceutical of Pharmacy, Faculty of Health and Medical Sciences, Sciences, Utrecht University (UU), David de Wiedgebouw, University of Copenhagen, Copenhagen, Denmark Universiteitsweg 99, 3584 CG Utrecht, The Netherlands 8 Agencia Espanola de Medicamentos y Productos Sanitarios 2 Julius Global Health, University Medical Center Utrecht (AEMPS), Madrid, Spain (UMCU), Utrecht, The Netherlands 9 Department of Clinical Pharmacy, UMCU, Utrecht, 3 Clinical Pharmacology Department, Hospital Universitari The Netherlands de la Vall d’Hebron, Universitat Autònoma de Barcelona 10 PHARMO Institute for Drug Outcomes Research, Utrecht, (UAB), Barcelona, Spain The Netherlands 4 Pharmacovigilance Research Center, Department of Drug 11 Department of Public Health, Maternal and Child Health, Design and Pharmacology, Faculty of Health and Medical Faculty of Medicine, Universidad Complutense de Madrid, Sciences, University of Copenhagen, Copenhagen, Denmark Madrid, Spain 5 Agenzia Regionale di Sanità della Toscana (ARS), Firenze, Italy 6 London School of Hygiene & Tropical Medicine (LSHTM), London, UK