The role of prenatal exposure to antidepressants, anxiolytic, and hypnotics and its underlying illness on the risk of miscarriage using BIFAP database.

PURPOSE: Despite the notable increase on the prescription of antidepressants and anxiolytics during pregnancy, recommendation on maintaining the treatment during prenatal period is still controversial. We aimed to separately assess the role of effects of the antidepressants and anxiolytic and the underlying illness, controlled by potential confounding associated with miscarriage onset. METHODS: We used data from a validated pregnant cohort aged 15-49 years from 2002 to 2016 using BIFAP database. All confirmed miscarriages were used to perform a nested control analysis using conditional logistic regression. Women were classified according to use of each drug of interest into four mutually exclusive groups: nonusers, users only during prepregnancy, continuers, and initiators during first trimester. Adjusted odds ratios (aORs) for major confounders during pregnancy such as number of visits to primary care practitioners visits, obesity, smoking, HTA, diabetes with 95% confidence intervals were calculated. RESULTS: Compared with nonusers, antidepressants continuers had the highest increased risk of miscarriage aOR (95%) of 1.29 (1.13-1.46), being continuers of paroxetine and fluoxetine the antidepressants with the strongest association. Likewise, continuers of anxiolytics and initiators showed an increased risk of 1.19 (1.04-1.37) and 1.30 (1.13-1.50). When separating the effect between the condition itself or the treatment, women exposed during first trimester, regardless treatment duration and/or the underlying illness, had the highest risk 1.27 (1.08-1.51) for antidepressants and 1.25 (1.13-1.39) for anxiolytics. CONCLUSIONS: Our analysis showed an association between prenatal exposure to antidepressants and anxiolytics and miscarriage onset after controlling by potential confounding adjusting for confounders and the underlying illness. This association was not supported for hypnotic medications. Further studies are warranted to evaluate the risk of miscarriage among subpopulation of pregnant women requiring these medications.

An algorithm to identify pregnancies in BIFAP Primary Care database in Spain: Results from a cohort of 155 419 pregnancies.

PURPOSE: There has been a notable progress on the development of methods for identification of pregnancies using primary care databases. We aimed to evaluate the prescription of medications during pregnancy applying a novel algorithm. METHODS: We identified pregnancies in women aged 15 to 49 years registered in the Database for Pharmacoepidemiological Research in Primary Care (BIFAP) between 2002 and 2015. The algorithm applied sequential cycles that searched in hierarchical order for indicators of conception, delivery or pregnancy loss, and other codes suggestive of pregnancy. Length of pregnancy was assessed by searching for last menstrual period (LMP) date, gestational age, and outcomes of pregnancy. Prescription of specific drugs during the pre-pregnancy period and first trimester and time trends during pregnancy were evaluated. RESULTS: We identified a total of 155 419 pregnancies during the study period (77.5% completed pregnancies, 21.5% pregnancies losses, 0.8% ectopic pregnancies, and 0.2% stillbirths). Excluding vitamins and supplements, 43.8% of women received at least one prescription during the pre-pregnancy period and 68.4% during the first trimester. During the first trimester, the most commonly drugs prescribed were analgesics (16.3%) followed by antibiotics (11.8%). From 2002/2003 to 2014/2015, there was an increase of prescriptions for thyroid hormones (1.0% vs 4.7%), H2 blockers (1.0% vs 2.2%), and PPIs (1.4% vs 2.2%). While antidepressants (2.0% vs 1.5%) and benzodiazepines (3.1% vs 2.4%) decreased in the last period. CONCLUSION: Having in mind the challenges of identifying pregnancies in health care databases, this study demonstrates the usefulness of BIFAP database for studies on drug utilization during pregnancy.

The recording of human papillomavirus (HPV) vaccination in BIFAP primary care database: A validation study.

PURPOSE: In Spain, a human papillomavirus (HPV) vaccine was firstly marketed in 2006 and mainly administered in primary care (PC) practices for girls/women or schools. As for all vaccines, a valid data source is required for research on observational effectiveness or safety. The objective of this study is to identify and validate HPV vaccinations recorded among women in The Primary Care Database For Pharmacoepidemiological Research (BIFAP) from 2007. METHODS: BIFAP includes a Vaccination File filled by PC practitionners and pediatricians. Information on women with HPV vaccinations recorded at any age was identified and validated according to whether (1) doses adhered to the following standard intervals: 27 to 269 days between first and second doses, >55 days between second and third doses, and <366 between first and third doses, and (2) additional information recorded in clinical records confirmed (through recording the brand, batch, expiring date, administration site, or vaccination comment) or refuted the vaccination and date. The latter was retrieved through manual review of anonymous records randomly selected. RESULTS: One hundred seventeen thousand seventy-three women with HPV vaccination records were identified (mean age 14.7 years); 82.5% had three jabs, 87.3% in recommended intervals. A sample of 978 patients' records, including 2245 jabs, was reviewed. Of the 363 jabs with additional information, 91% confirmed the vaccination. Confirmatory data was more frequent when doses strictly adhered to recommendations (96.8%-100%) than not (60.0%-85.7%). CONCLUSIONS: In BIFAP, a cohort of women vaccinated against HPV, mostly with three doses in recommended intervals, was identified. Although additional information about the vaccination was scarce, when present, it highly confirmed it, making BIFAP a potential data source for HPV vaccine research.

Impact of the 2018 revised Pregnancy Prevention Programme by the European Medicines Agency on the use of oral retinoids in females of childbearing age in Denmark, Italy, Netherlands, and Spain: an interrupted time series analysis.

Background: In March 2018, the European pregnancy prevention programme for oral retinoids was updated as part of risk minimisation measures (RMM), emphasising their contraindication in pregnant women. Objective: To measure the impact of the 2018 revision of the RMMs in Europe by assessing the utilisation patterns of isotretinoin, alitretinoin and acitretin, contraceptive measures, pregnancy testing, discontinuation, and pregnancy occurrence concomitantly with a retinoid prescription. Methods: An interrupted time series (ITS) analysis to compare level and trend changes after the risk minimisation measures implementation was conducted on a cohort of females of childbearing age (12-55 years of age) from January 2010 to December 2020, derived from six electronic health data sources in four countries: Denmark, Netherlands, Spain, and Italy. Monthly utilisation figures (incidence rates [IR], prevalence rates [PR] and proportions) of oral retinoids were calculated, as well as discontinuation rates, contraception coverage, pregnancy testing, and rates of exposed pregnancies to oral retinoids, before and after the 2018 RMMs. Results: From 10,714,182 females of child-bearing age, 88,992 used an oral retinoid at any point during the study period (mean age 18.9-22.2 years old). We found non-significant level and trend changes in incidence or prevalence of retinoid use in females of child-bearing age after the 2018 RMMs. The reason of discontinuation was unknown in >95% of cases. Contraception use showed a significant increase trend in Spain; for other databases this information was limited. Pregnancy testing was hardly recorded thus was not possible to model ITS analyses. After the 2018 RMM, rates of pregnancy occurrence during retinoid use, and start of a retinoid during a pregnancy varied from 0.0 to 0.4, and from 0.2 to 0.8, respectively. Conclusion: This study shows a limited impact of the 2018 RMMs on oral retinoids utilisation patterns among females of child-bearing age in four European countries. Pregnancies still occur during retinoid use, and oral retinoids are still prescribed to pregnant women. Contraception and pregnancy testing information was limited in most databases. Regulators, policymakers, prescribers, and researchers must rethink implementation strategies to avoid any pregnancy becoming temporarily related to retinoid use.

Assessment of thyroiditis risk associated with HPV vaccination among girls aged 9-18 years: A time-varying cohort study.

BACKGROUND: Previous studies have suggested a relationship between human papillomavirus vaccine and autoimmune diseases, including thyroiditis. Thus, we aimed to evaluate the risk of thyroiditis associated with HPV vaccination among girls using the Primary Care Database For Pharmacoepidemiological Research (BIFAP) in Spain. METHODS: In this retrospective cohort study, girls in BIFAP aged 9-18 years from 2007 to 2016, free of past thyroiditis and HPV vaccination, were included. Hazard Ratios (HRs; 95% CI) of thyroiditis were calculated within exposed periods (up to 2 years of vaccination) and post-exposed periods (from 2 years after vaccination onwards) compared with non-exposed periods, overall, by dose and by type of vaccine, adjusted for potential confounders collected at different times. In a post-hoc analysis, we moved back the thyroiditis date (30 days) as a theoretical delay in diagnosis. RESULTS: Out of the 388,411 girls included in the cohort, 153,924 were vaccinated against HPV and 480 thyroiditis (253 autoimmune) cases were identified (334 non-exposed; 103 exposed; 43 post-exposed). Adjusted HR was 1.18 [95% CI: 0.79-1.76] for exposed (1.25 [0.77-2.04] for bi- and 1.15 [0.76-1.76] for quadri-valent vaccines) and 1.26 [0.74-2.14] for post-exposed periods. HR was 1.50 [0.87-2.59] for the 1st dose, 1.13 [0.66-1.91] for the 2nd and 1.11 [0.71-1.72] for the 3rd one. When the diagnosis date was moved back, the risk was 1.14 [0.76-1.70] for exposed period, being 1.80 [0.86-3.76] and 1.40 [0.74-2.66] after 1st dose of bi- and quadri-valent, respectively. CONCLUSIONS: We did not observe an increased risk of thyroiditis following HPV vaccination (whether bi- or quadri-valent). Even though the point estimate was higher after 1st HPV vaccination dose than after subsequent doses, a dose-effect was not confirmed. Results remained similar after applying a lag time.

Flecainide and risk of skin neoplasms: Results of a large nested case-control study in Spain and Denmark.

Background: A previous study in Denmark suggested an increased melanoma risk associated with the use of flecainide. Objective: To study the association between flecainide use and the risk of melanoma and non-melanoma skin cancer in Spain and Denmark. Methods: We conducted a multi-database case-control study in (database/study period) Spain (SIDIAP/2005-2017 and BIFAP/2007-2017) and Denmark (Danish registries/2001-2018). We included incident cases of melanoma or non-melanoma skin cancer (NMSC) aged ≥18 with ≥2 years of previous data (≥10 years for Denmark) before the skin cancer and matched them to controls (10:1 by age and sex). We excluded persons with immunosuppression or previous cancer. We defined ever-use as any prescription fill and high-use as a cumulative dose of at least 200 g (reference: never-use). We categorized a cumulative dose for a dose-response assessment. We used conditional logistic regression to compute ORs (95% CI) adjusted for photosensitizing, anti-neoplastic, disease-specific drugs and comorbidities. Results: The total numbers of melanoma/NMSC cases included were 7,809/64,230 in SIDIAP, 4,661/31,063 in BIFAP, and 27,978/152,821 in Denmark. In Denmark, high-use of flecainide was associated with increased adjusted ORs of skin cancer compared with never-use [melanoma: OR 1.97 (1.38-2.81); NMSC: OR 1.34 (1.15-1.56)]. In Spain, an association between high-use of flecainide and NMSC was also observed [BIFAP: OR 1.42 (1.04-1.93); SIDIAP: OR 1.19 (0.95-1.48)]. There was a non-significant dose-response pattern for melanoma in Denmark and no apparent dose-response pattern for NMSC in any of the three databases. We found similar results for ever-use of flecainide. Conclusion: Flecainide use was associated with an increased risk of melanoma (Denmark only) and NMSC (Denmark and Spain) but without substantial evidence of dose-response patterns. Further studies are needed to assess for possible unmeasured confounders.

Papillomavirus vaccination and Guillain-Barre Syndrome among girls: A cohort study in Spain.

BACKGROUND: Studies of the association of Guillain-Barré Syndrome (GBS) with papillomavirus vaccination (HPVv; scheduled from 2007) have provided contradicting results, probably due to the low frequency of this disease. We aimed at estimating that risk relative to non-vaccination among girls, by using the Spanish Primary Care Database for Pharmacoepidemiological Research (BIFAP). METHODS: A cohort study of girls aged 9-18 years during 2007-2016 free of GBS or HPVv was selected and followed up to GBS diagnosis. Follow-up time was divided by time-varying HPVv exposure and confounders. Crude Incidence rates (IR per 1,000,000 person-years (py)) and adjusted Hazard Ratios (HR) of GBS were estimated anytime after vaccination compared to non-exposed periods. HRs were also estimated for the first 90 days after HPVv (risk-window) and thereafter. RESULTS: Out of 388,849 girls, of which 154,255 were vaccinated, 6 'confirmed' GBS cases occurred during non-exposure periods (IR of 5.83 per million person-years; 95% CI: 2.62-12.97) and 3 'confirmed' cases anytime after vaccination (IR of 7.87; 95% CI: 2.54-24.39). The resulting adjusted HR anytime after vaccination was 1.24 (95% CI: 0.19-8.00). All three cases occurred after the risk window of 90 days with an HR of 1.77 (95% CI: 0.25-12.54) for post-exposure periods as compared with non-exposure. Since zero cases occurred during the risk window, no HR could be estimated for exposed periods. CONCLUSIONS: Incidences of GBS were in line with the range previously reported for young people, supporting the potential of BIFAP for performing studies on GBS. However, a lack of power may be present for quantifying the relative risk of such a rare disease after the vaccination among the study cohort, where we can only exclude an increased risk of 8-times relative to no vaccination.

A Validation Study on the Frequency and Natural History of Miscarriages Using the Spanish Primary Care Database BIFAP.

(1) Background: There is a major gap of knowledge towards the natural history of miscarriages in electronic medical records. We aimed to calculate the frequency of miscarriages using data from BIFAP database. (2) Methods: We identified all pregnancy losses and carried out a multistep validation exercise. Potential cases with positive predictive values (PPV) of miscarriage confirmation <85% or those confirming other pregnancy loss were excluded. Kaplan-Meier figures and incidence rates (IRs) of miscarriage with 95% confidence intervals (CIs) expressed by 1000 person-weeks were calculated. Stratifying analysis by age, specific high-risk groups, and drug exposure within the pre-pregnancy period were performed restricted to women with recording last menstrual period (LMP). (3) Results: Women with confirmed miscarriage (N = 18,070), tended to be older, with higher frequency of comorbidities and drug utilization. Restricting to women with LPM recorded, IR of miscarriage was 10.89 (CI 95% 10.68-11.10) per 1000 women-weeks, with a median follow-up of 10 weeks (IQR: 8-12). The IR according to age was: 2.71 (CI 95% 2.59-2.84) in those aged <30 years compared to 9.11 (CI 95% 8.55-9.70) in women aged ≥40 years. Advanced maternal age (Hazard Ratio (HR, 95% confidence interval) CI 95%: 3.34 (3.08-3.62)), use of antihypertensives (1.49 (1.21-1.84), and use of drugs classified as D or X during pregnancy (1.17 (1.07-1.29)) showed to be positive predictors associated with increased risk of miscarriages. (4) Conclusion: BIFAP database can be used to identify women suffering from miscarriages, which will serve to further study risk factors associated with miscarriages with special attention to drug utilization.

Evaluation of the Risk of Inflammatory Bowel Disease after the HPV Vaccination in Primary Care in Spain: A Time-Varying Cohort Analysis of Around 390,000 Girls.

INTRODUCTION: A link between the human papillomavirus vaccination (HPVv) and inflammatory bowel disease (IBD) has been suggested. OBJECTIVE: We aimed to estimate the risk of IBD following HPVv compared with periods not exposed to the vaccines. METHODS: Primary healthcare records (Spanish Primary Care Database For Pharmacoepidemiological Research [BIFAP]) were used in a cohort study of girls in Spain aged 9-18 years between 2007 and 2016 free of IBD or HPVv at study entrance. During the follow-up to IBD diagnosis, time-varying HPVv exposure and confounders were assessed in Cox models to estimate the hazard ratio (HRs) of IBD in the 2 years after HPVv (exposed period) and thereafter (post-exposed) compared with the no exposure periods. In a post hoc analysis, we moved the IBD date back 30 days as a theoretical delay in diagnosis confirmation. RESULTS: The cohort comprised 388,669 girls; 154,174 of these received the HPVv, and 88 IBD cases occurred (55 non-exposed, 22 exposed [after first N = 6, second N = 2, or third N = 14 dose] and 11 in post-exposed periods). The adjusted HR was 1.66 (95% confidence interval [CI] 0.68-4.05) for exposed and 1.10 (95% CI 0.37-3.24) for post-exposed periods. The HR for the first dose was 3.94 (95% CI 1.19-13.02). No association was found for the second or third doses. Post hoc, the HR was 1.83 (95% CI 0.72-4.69) for exposed periods (N = 18), and 1.84 (95% CI 0.35-9.83; N = 2), 1.50 (95% CI 0.40-5.63; N = 4) and 1.98 (95% CI 0.71-5.49; N = 12) after the first, second and third doses, respectively. CONCLUSIONS: This study did not show an increased risk of IBD following 2 years of HPVv exposure. However, an increased risk of IBD diagnosis was observed following the first vaccination dose (1-34 days), which is likely attributable to the clinical recommendation to vaccinate upon onset of IBD symptoms.

Age-specific vaccination coverage estimates for influenza, human papillomavirus and measles containing vaccines from seven population-based healthcare databases from four EU countries - The ADVANCE project.

BACKGROUND: The Accelerated Development of VAccine beNefit-risk Collaboration in Europe (ADVANCE) is a public-private collaboration aiming to develop and test a system for rapid benefit-risk monitoring of vaccines using existing healthcare databases in Europe. We estimated vaccine coverage from electronic healthcare databases as part of a fit-for-purpose assessment for vaccine benefit-risk studies. METHODS: A retrospective dynamic cohort study was conducted through a distributed network approach. Coverage with measles-vaccine for birth year 2006, human papillomavirus (HPV)-vaccine for birth years 1990-2000 and influenza-vaccine for birth years 1920-1950 was estimated using period-prevalence and inverse probability weighting methods. Seven databases from four countries participated: Italy (Pedianet, Val Padana), Spain (BIFAP, SIDIAP), UK (RCGP-RSC, THIN), Denmark (SSI/AUH). Database access providers extracted the data, transformed it into a common structure and ran an R-script locally. The created output tables were shared and pooled at a central server. RESULTS: The total study population comprised 274,616 persons for measles-vaccine, 2,011,666 persons for HPV-vaccine and 14,904,033 persons for influenza-vaccine. Measles-vaccine coverage varied from 84.3% (Denmark) to 96.5% (Italy, Val Padana) for the first dose and from 82.8% (Italy, Val Padana) to 90.9% (UK) for the second dose at the age of 7 years. The HPV-vaccine coverage, aggregated over birth years 1997-2000, ranged from 60% (UK) to 88.3% (Denmark) at the age of 15 years. The influenza-vaccine coverage for the influenza seasons from 2009 to 2015 for persons aged 65 years and more was roughly stable around 43% in Denmark and around 68% in the UK while a decrease from 58 to 50% was observed in Catalonia (Spain). CONCLUSIONS: We obtained detailed, age-specific coverage estimates though a common procedure. We discussed between database comparability and comparability to published national estimates.

RETRACTED ARTICLE: The Longitudinal Incidence of Human Papillomavirus Vaccination in Spanish Primary Care in the First 6 Years After Approval.

BACKGROUND: In Spain, girls and women are vaccinated against human papillomavirus (HPV) in the primary care setting, according to a national vaccination program. Vaccination is voluntary and the cost is covered by the public health system. OBJECTIVES: The aim of the study was to estimate the incidence and patterns of HPV vaccination amongst girls in Spain. METHODS: A cohort study was performed using the information recorded in the Spanish Primary Care Database for Pharmacoepidemiological Research (BIFAP) from 7.4 million patients from eight Spanish regions, between 2001 and 2013 (56% of the regional population). Data available in BIFAP include patient age, sex, lifestyle factors, clinical events, specialist referrals, prescriptions, and vaccinations as recorded by the primary care physician (PCP) or administering nurse. The study cohort comprised all girls aged 11-18 years registered in BIFAP between 1 January 2007 and 31 December 2013 who had at least 1 year of clinical record information with their PCP (inclusion criteria). The date the inclusion criteria were met was designated as the start date of the study cohort contribution. In order to estimate the incidence of HPV vaccination, girls forming the study cohort were followed from start date until there was a recorded HPV vaccination, they reached 19 years of age or died, the end of available information, or 31 December 2013. The person-time of all patients forming the study cohort was taken into account in the incidence estimations. The cumulative incidence (CuIn) of vaccination by birth cohort, year and region was estimated using life-tables (proportion of vaccination by intervals in which the denominator is the initial population corrected for losses). RESULTS: Out of 273,098 girls forming the study population, 81,461 were vaccinated during 2007-2013. Age ranged from 12 to 14 years at first dose in 86.0% of vaccinated girls; 54.1% received a quadrivalent vaccine, 21.9% a bivalent vaccine, and 24.0% an unknown type. Out of the vaccinated population, 87.9% received three doses, 8.2% two and 3.9% one dose, at a maximum of 7 years of follow-up. By calendar year and region, the CuIn reached 70.0-95.8% for birth cohorts between 1993 and 1999, 28.6-99.0% for births cohorts between 1990 and 1992, and exceptionally, 70.6-99.8% for births cohorts between 2000 and 2002 in three regions. CONCLUSIONS: According to BIFAP primary care data, a high incidence of vaccination among girls aged 13-15 years was observed. Vaccination among younger and older girls was less common although they reached high incidence in those regions that included girls aged 11-18 years in mass programs. Most vaccination patterns adjusted to a three-dose regimen, as recommended.

The Longitudinal Incidence of Human Papillomavirus Vaccination in Spanish Primary Care in the First 10 Years After Approval.

BACKGROUND: In Spain, girls are vaccinated against human papillomavirus (HPV) in the primary care setting, according to a national vaccination programme. Vaccination is voluntary and is covered by the public health system. OBJECTIVES: The aim of the study was to estimate the incidence and patterns of HPV vaccination amongst girls in primary care in Spain. METHODS: A cohort study was performed using the information recorded in the Spanish Primary Care Database for Pharmacoepidemiological Research (BIFAP) from 7.9 million patients from seven Spanish regions, between 2001 and 2016 (56.6% of the regional population). Data available in BIFAP include patient age, sex, life-style factors, clinical events, specialist referrals, prescriptions, and vaccinations as recorded by the primary care physician (PCP) or administering nurse. The study cohort comprised all girls aged 9-18 years registered in BIFAP between 1st January of 2007 and 31st December of 2016 who had at least 1 year of clinical record information with their PCP (inclusion criteria). The date the inclusion criteria were met was designated as the start date to the study cohort contribution. In order to estimate the incidence of HPV vaccination (initiation of vaccination schedule), girls with an HPV vaccination recorded before the start date or without vaccination date were excluded. Girls forming the study cohort were followed from start date until there was a recorded HPV vaccination, they reached 19 years of age or died, end of available information, or 31st December 2016. The person-time of all patients forming the study cohort was reckoned in the incidence estimations. The follow-up was replicated yearly from 2007 to 2016. The cumulative incidence (CuIn) of vaccination by birth cohort, year and region, was estimated using life tables (proportion of vaccination by intervals in which the denominator is the initial population corrected for losses). RESULTS: Of 388,690 girls forming the study population, 154,211 initiated the vaccination during 2007-2016. Ages ranged from 12 to 14 years at first dose in 84.5% of vaccinated girls, 42.79% received a quadrivalent vaccine, 21.86% a bivalent vaccine, and 35.35% an unknown type. Of the vaccinated population, 48.0% were completely vaccinated with a three-dose schedule and 28.9% with a two-dose schedule, 20.2% received one dose and 3.0% two doses in a three-dose schedule, at a maximum of 10 years of follow-up. The CuIn was highest among girls aged either 13 or 14 years over all regions (reaching 92.8% and 89.7%, respectively), and aged 12 in some regions/years (up to 89.8%). Girls aged 15 years were also vaccinated (although showing lower yearly incidence, i.e. < 69.1%) in two regions. The coverage was broadened to younger girls (11 years) during the last years of the study period in some regions. CONCLUSIONS: According to BIFAP primary care data, a high incidence of vaccination among girls aged 13-14 years was observed. Vaccination among younger and older girls were less common, although they reached high incidence in some regions and/or years. Most vaccination patterns adjusted to a complete vaccination regimen, as recommended posology.

The impact of different strategies to handle missing data on both precision and bias in a drug safety study: a multidatabase multinational population-based cohort study.

BACKGROUND: Missing data are often an issue in electronic medical records (EMRs) research. However, there are many ways that people deal with missing data in drug safety studies. AIM: To compare the risk estimates resulting from different strategies for the handling of missing data in the study of venous thromboembolism (VTE) risk associated with antiosteoporotic medications (AOM). METHODS: New users of AOM (alendronic acid, other bisphosphonates, strontium ranelate, selective estrogen receptor modulators, teriparatide, or denosumab) aged ≥50 years during 1998-2014 were identified in two Spanish (the Base de datos para la Investigación Farmacoepidemiológica en Atención Primaria [BIFAP] and EpiChron cohort) and one UK (Clinical Practice Research Datalink [CPRD]) EMR. Hazard ratios (HRs) according to AOM (with alendronic acid as reference) were calculated adjusting for VTE risk factors, body mass index (that was missing in 61% of patients included in the three databases), and smoking (that was missing in 23% of patients) in the year of AOM therapy initiation. HRs and standard errors obtained using cross-sectional multiple imputation (MI) (reference method) were compared to complete case (CC) analysis - using only patients with complete data - and longitudinal MI - adding to the cross-sectional MI model the body mass index/smoking values as recorded in the year before and after therapy initiation. RESULTS: Overall, 422/95,057 (0.4%), 19/12,688 (0.1%), and 2,051/161,202 (1.3%) VTE cases/participants were seen in BIFAP, EpiChron, and CPRD, respectively. HRs moved from 100.00% underestimation to 40.31% overestimation in CC compared with cross-sectional MI, while longitudinal MI methods provided similar risk estimates compared with cross-sectional MI. Precision for HR improved in cross-sectional MI versus CC by up to 160.28%, while longitudinal MI improved precision (compared with cross-sectional) only minimally (up to 0.80%). CONCLUSION: CC may substantially affect relative risk estimation in EMR-based drug safety studies, since missing data are not often completely at random. Little improvement was seen in these data in terms of power with the inclusion of longitudinal MI compared with cross-sectional MI. The strategy for handling missing data in drug safety studies can have a large impact on both risk estimates and precision.