Decision tree analysis, a flowchart-like tree framework, is a typical machine learning method that is widely used in various fields. The most significant feature of this method is that independent variables (e.g., with or without concomitant use of vasopressor drugs) are extracted in order of the strength of their relationship with the dependent variable to be predicted (e.g., with or without adverse drug reactions), forming a tree-like model. Specifically, users can easily and quantitatively estimate the proportion of event occurrences considering “interrelationships among multiple combinations of factors” by answering the questions in the constructed flowchart. Previously, we applied the decision tree model to vancomycin-associated nephrotoxicity and demonstrated that this method can be used to analyze the factors affecting adverse drug reactions. However, the number of cases that can be analyzed decreases significantly as the number of branches increases. Thus, many cases are necessary to generate highly accurate findings. In attempt to solve this problem, we combined big data and decision tree analyses. In this review, we present the results of our research combining big data (electronic medical record database) and a machine learning method. Furthermore, we discuss the limitations of these methods and factors to consider when applying the results of big data and machine learning analyses to clinical practice.

Recent developments have enabled daily accumulated medical information to be converted into medical big data, and new evidence is expected to be created using databases and various open data sources. Database research using medical big data was actively conducted in the coronavirus disease 2019 (COVID-19) pandemic and created evidence for a new disease. Conversely, the new term “infodemic” has emerged and has become a social problem. Multiple posts on social networking services (SNS) overly stirred up safety concerns about the COVID-19 vaccines based on the analysis results of the Vaccine Adverse Event Reporting System (VAERS). Medical experts on SNS have attempted to correct these misunderstandings. Incidents where research papers about the COVID-19 treatment using medical big data were retracted due to the lack of reliability of the database also occurred. These topics of appropriate interpretation of results using spontaneous reporting databases and ensuring the reliability of databases are not new issues that emerged during the COVID-19 pandemic but issues that were present before. Thus, literacy regarding medical big data has become increasingly important. Research related to artificial intelligence (AI) is also progressing rapidly. Using medical big data is expected to accelerate AI development. However, as medical AI does not resolve all clinical setting problems, we also need to improve our medical AI literacy.

With the development of information technology, patient information is stored as electronic data, and huge amounts of such data are collected every day. Such a collection compiled over the course of clinical practice is called real-world data and is expected to be used for evaluating drug efficacy and safety. Real-world data such as health insurance association-based administrative claims databases, pharmacy-based dispensing databases, and spontaneous reporting system databases are mainly used in pharmaceutical research. Among them, claims databases are used for various observational studies such as studies on nationwide prescription trends, pharmacovigilance studies, and studies on rare diseases due to their large sample size. Although the nature of omics data is different from that of real-world data, it has become accessible on cloud platforms and are being used to broaden the scope of research in recent years. In this paper, we introduce a method for generating and further testing hypotheses through integrated analysis of real-world data and omics data, with a focus on administrative claims databases.

Industrial reforms utilizing artificial intelligence (AI) have been progressing remarkably worldwide in recent years. In medical informatics, medical big-data analytics involving AI are increasingly being promoted, and AI in the medical field is being widely applied in research areas such as protein-structure analysis and diagnostic support. Previously, we developed a unique adverse drug reactions analysis system that incorporates Accord.NET, an open-source machine learning (ML) framework written in the programming language C#, and uses the Japanese Adverse Drug Event Report (JADER) database. The developed system can provide necessary information for exploratory investigation of drug efficacy, side effects, adherence, and so on. To efficiently interpret the calculated data and minimize noise, the developed system features a data visualization tool that can visualize the results of various statistical analyses and machine learning models in real-time three dimensions (3D), making it intuitive to grasp the results. This feature makes the system ideal for individuals in clinical work. We believe that the system will facilitate more efficient drug management and clinical pharmacy research. In this review, we introduce an example of domain-driven design development of this AI analysis system for pharmacists in clinical practice with the aim of further utilizing medical big data and AI analytics.

Medical big data is accumulated numerous medical related data day by day. These data may have tips for new approach for drug development. Authors tried to find drug-development-needs in children using medical big data analysis with prescription survey. Medical big data were provided from JMDC (Japan Medical Data Centre) Inc. about 3 million participants between January 2005 and June 2017. In these, we identified randomly identified 22787 participants from 466701 participants who are aged ≤11 years. In these participants, 9644 were administered “capsule,” “tablet,” “orally distegrating tablet,” “controlled release tablet/capsule” or “enteric coated tablet” formula drugs. In these, 514 were administered these as powderization or decapsulation. Sixty components administered in 145 participants (28.2%) are not marketed for pediatric formula. On the other hands, 92 components administered in 369 participants (71.8%) are decapsulation or powderization, though pediatric formulas are marketed. These 152 components may have a development seeds for children. In conclusion, prescription survey using medical big data may partially resolve the drug-development-need in pediatrics because by using medical big data will leads low biased data depending each institution.

Unlike many researchers of natural product chemistry, I was fortunate to have the opportunity to study phytochemical metabolites and isolates of microbial origin. I began my career isolating the active compound(s) from the medicinal plants. After obtaining a Ph.D. degree from Tohoku University, I flew to Chicago, College of Pharmacy, University of Illinois, where I carried out research on the chemistry of acronycine and discovered several interesting chemical reactions regarding this alkaloid. After returning to Japan, I began to work at the Kitasato Institute, to search for novel antitumor antibiotics. During this period, 27 new antibiotics were isolated, and the new chemical structures were elucidated. After rejoining the Pharmaceutical Institute at Tohoku University, I again began to work on the phytochemical substances, mainly alkaloids. These studies continued after I moved to Aomori University and finally to Nihon Pharmaceutical University. I was interested in the biosynthesis of the alkaloids and found that all alkaloids could be classified into 16 classes based on their method of biosynthesis. I wrote a book about this in Japanese, and subsequently the book was translated into English as “ALKALOIDS—A Treasury of Poisons and Medicines.” After completing the publication of this book, I had many chances to write books, mainly concerning poisons and medicines. Totally, I have been able to publish 26 books regarding on these fascinating topics until now. I am feeling very satisfied with my natural product chemistry contributions, especially those of alkaloids and poisons.

To accelerate therapeutic effects, the mixtures of two or more topical pharmaceutical products having different medicinal purposes are often applied in the medical field. In this study, we aimed to develop a simple mixing method/procedure to achieve excellent homogeneity in the mixture of two topical products, a steroidal ointment and a skin moisturizer. To assess an in-tube mixing method as a simple mixing procedure, we injected both topical products into an empty resin tube, a flexible hollow tube with an open end that can be closed on one side, and a closed end on the other, removed as less air as possible inside the tube, and then thermocompressed (sealed) the open end to close it. The two topical products were then mixed uniformly by repeated finger pressure along the longitudinal axis of the tube. The homogeneity of the two topical products in the tube was evaluated by measuring the content of methyl paraoxybenzoate (MP), an additive loaded in the skin moisturizer. In addition, the mixability was qualitatively evaluated from the distribution of white petrolatum, another additive loaded in the steroid ointment, using Raman spectroscopy. As a result, the measured value of MP relative to the label claim was in the range of 100±12%, and the coefficients of variation value was also less than 12%. These results indicate that the in-tube mixing method using two topical products is approximately hologenetic preparations that do not cause therapeutic problems.

A 72-year-old man with a malignant retroperitoneal soft tissue tumor was treated with ifosfamide (IFO) for 5 consecutive days (1.8 g/m²/d×5 d, expected dose 9 g/m²). The patient developed neurological symptoms such as mild somnolence, seizures, and inability to write from Day 1, and became delirious on Day 3, so IFO was discontinued on Day 4 (dose: 7.2 g/m²). Since there are reports of drug interactions that increase the frequency of encephalopathy when combined with aprepitant (Apr), Dexamethasone was increased and IFO was administered without the use of Apr after the second course, and there was no recurrence of encephalopathy in the second and third courses. IFO-induced encephalopathy is considered to occur due to an increase in blood concentration of IFOs caused by high dosage, decreased renal function, or other factors. In this case, encephalopathy was observed even though the dose of IFO was reduced due to the patient’s advanced age and impaired renal function. The combination use of Apr with IFO should be considered with caution for the occurrence of adverse events, including encephalopathy, and if possible, control of gastrointestinal toxicity with other antiemetic agents should be considered.

Therapeutic drug monitoring (TDM) is recommended for voriconazole (VRCZ) to avoid adverse events and maximize antifungal efficacy. Currently, the appropriate dose for patients under the age of 2 years is unknown. Here, we report the case of a 1.5-month-old infant with inborn errors of immunity who was orally administered VRCZ. This patient’s plasma concentration decreased significantly from 3.8 µg/mL (day 6) to 0.09 µg/mL (day 21), leading to repeated dose escalations to achieve the target concentration (1.38 µg/mL, day 58). The signal intensity ratio of VRCZ to its main metabolite, N-oxide VRCZ, in LC/MS/MS also decreased from 5.30 (day 6) to 0.57 (day 64). Consequently, we suspected that VRCZ metabolism may be enhanced during infant growth. To our knowledge, this is the first report of remarkable changes in VRCZ pharmacokinetics with metabolic activity enhanced by the growth process. In conclusion, we propose that frequent TDM helped to maintain adequate VRCZ plasma concentration in a infants less than 6 months of age.