Epidemiological methods have been applied to investigate drug problems such as past drug disasters, and the academic field called pharmacoepidemiology was created. The first international conference of pharmacoepidemiology was held in 1985, and the first Japanese conference was in 1995. Therefore it is the relatively new field. Recently, pharmacoepidemiology has gained a lot of attention because of US sentinel initiative, recommendations by the Ministry of Health, Labor, and Welfare in Japan, and revision of GPSP for analyzing medical databases with epidemiological methods. In the future of pharmacoepidemiology, it is expected that the quality and quantity improvements of medical databases, and signal detection based on IoX and AI innovation. In addition, genomic data will be also more available and pharmacoepidemiology gets much closer to genomic epidemiology. It would be also possible to linkage between clinical data and patient registries, and improve analytical methods. Also, I would like to hope that pharmacoepidemiology gets more attention due to not merely big data, but creating knowledge on the safety of medicines.

Although the recent revision of the ministerial ordinance on Good Post-marketing Study Practice (GPSP) included the utilization of medical information databases for post-marketing surveillance, there has been limited research on the validity of diagnosis codes and other outcome definitions in Japanese databases such as administrative claims (“receipt”) database. This task force proposed how to conduct good validations studies, based on the narrative review on around 100 published papers around the world. The established check list consists of ： (ⅰ) understanding the type of the database (e.g. administrative claims data, electronic health records, disease registry) ； (ii) understanding the setting of the validation study (e.g. “population-based” or not) ； (iii) defining the study outcome ； (iv) determining the way of linkage between databases ； (v) defining the gold standard ； (vi) selecting the sampling method (e.g. using the information of all patients in the database or a hospital, random sampling from all patients, random sampling from patients satisfying the outcome definition, random sampling from patients satisfying and not satisfying the outcome definition, “all possible cases” method) and sample size ； (vii) calculating the measures of validity (e.g. sensitivity, specificity, positive predictive value, negative predictive value) ； and (viii) discussing how to use the result for future studies. In current Japan, where the linkage between databases is logistically and legally difficult, most validation studies would to be conducted on a hospital basis. In such a situation, detailed description of hospital and patient characteristics is important to discuss the generalizability of the validation study result to the entire database. This report is expected to encourage and help to conduct appropriate validation studies.

Objective：The objective of this study was to apply Least Absolute Shrinkage and Selection Operator （LASSO）logistic regression to detection of adverse drug reaction （ADR） signals using an electronic health records database as a comprehensive and quantitative method to supplement the current pharmacovigilance activities in Japan.

Design：case-control study

Methods：We analyzed data from 40767 inpatients using a single-institution hospital database and identified two ADRs, suspected pancreatitis and thrombocytopenia, using abnormal laboratory test results. LASSO logistic regression analysis was applied to detect ADR signals with adjustment for age, sex, comorbidities and medical procedures. The positive predictive value （PPV） was calculated using reference standard of known drug-ADR associations based on drug product labels.

Results：The number of case group was 6735 for suspected pancreatitis and 11561 for thrombocytopenia. The number of ADR signals detected using LASSO logistic regression was 27 for suspected pancreatitis and 40 for thrombocytopenia. The calculated PPV was 3.7% for suspected pancreatitis and 55.0% for thrombocytopenia.

Conclusion：LASSO logistic regression analysis efficiently detects ADR signals by adjusting for confounding factors such as comorbidities and medical procedures. The false positive signals may contain unknown signals and further signal assessment will be needed.

Objective：The objective of this study was to apply Least Absolute Shrinkage and Selection Operator （LASSO）logistic regression to detection of adverse drug reaction （ADR） signals using an electronic health records database as a comprehensive and quantitative method to supplement the current pharmacovigilance activities in Japan.Design：case-control studyMethods：We analyzed data from 40767 inpatients using a single-institution hospital database and identified two ADRs, suspected pancreatitis and thrombocytopenia, using abnormal laboratory test results. LASSO logistic regression analysis was applied to detect ADR signals with adjustment for age, sex, comorbidities and medical procedures. The positive predictive value （PPV） was calculated using reference standard of known drug-ADR associations based on drug product labels.Results：The number of case group was 6735 for suspected pancreatitis and 11561 for thrombocytopenia. The number of ADR signals detected using LASSO logistic regression was 27 for suspected pancreatitis and 40 for thrombocytopenia. The calculated PPV was 3.7% for suspected pancreatitis and 55.0% for thrombocytopenia.Conclusion：LASSO logistic regression analysis efficiently detects ADR signals by adjusting for confounding factors such as comorbidities and medical procedures. The false positive signals may contain unknown signals and further signal assessment will be needed.

The notification of RMP was released in 2012 and has been adapted for new drug submission since 2013. However, most cases are usual post-marketing surveillance studies. According to the ICH E2E guideline, various risk managements could be possible, especially using medical database. Recently, large database has been developed. There are two kinds of database, hospital information system including electronic medical records, and claim data. Activities of using medical database in Japan, US, and Europe are various. Based on FDA amendment acts, FDA launched Sentinel Initiative in 2008 and REMS works effectively. The Mini-Sentinel and OMOP published Common Data Model respectively. FDA also released guidance for pharmacoepidemiologic studies using electronic health data. In Europe, RMP has been implemented in 2005 and about 36% are epidemiologic studies. ENCePP which was established in 2006 provides register of pharmacoepidemiologic and pharmacovigilance studies, checklist for protocols and guide on methodological standards in pharmacoepidemiology. In Japan, PMDA provides guideline for pharmacoepidemiologic studies using medical database. Also, “MID-NET” which is the standardized medical database has been developed. As a notable activity, PMDA has conducted pilots as MIHARI project and itʼs quite promising.