Real-world data (RWD), such as a health insurance claim database and electronic medical record database, which records daily medical care information, is one of the most important data sources in pharmacoepidemiological studies. In Japan, a post-marketing database study has been added as a new category of post-marketing surveillance since April 2018. It is expected that the post-marketing database studies will be conducted in the actual risk management plan； however, at this point, few have reached the stage of publishing study results. On the other hand, overseas, many database research results using RWD have already been reported. Although it is necessary to keep in mind that there are differences in the characteristics of the databases (e.g., data contents, structures) and the differences in the medical environment between overseas and Japan, we think that critically reviewing these reports has a reference value for the planning, execution, and interpretation of the results in performing post-marketing database studies in Japan. The purpose of this report is to examine the characteristics and cautions of the database studies through a critical review of published articles of overseas RWD studies and to make recommendations that are useful for conducting post-marketing database studies in Japan. We hope this article will become a help in the planning and implementation of future post-marketing database studies.

Objectives: To evaluate the feasibility and efficacy of treatment for the recovery of knee joint function after total knee arthroplasty (TKA) using a robotic suit.Patients and Methods: Knee joint extension exercise sessions were started with a robotic suit (single-joint hybrid assistive limb [HAL-SJ, Cyberdyne, Inc., Tsukuba, Japan]) in one group of patients after TKA. Patients who underwent standard rehabilitation were enrolled in the control group. To evaluate feasibility and safety, we assessed the adverse events, the number of training sessions, and training time. We compared the changes in knee joint pain and extension lag (°) between the groups.Results: The average age was 71.3 ± 6.2 years in the HAL-SJ group and 74.9 ± 8.7 years in the control group. There were no severe adverse events. In the HAL-SJ group, training was performed 2.9 times, on average, and lasted 18.8 min. In the HAL-SJ group, there was a reduction in the visual analog scale (VAS) for pain after training, which was not significant. In the control group, the VAS score worsened after the sessions. The extension lag significantly improved in the HAL-SJ group after the 2nd and 3rd sessions, and this was more due to improvements in their active extension range of motion than their passive extension range of motion.Conclusions: HAL-SJ-based training is safe and effective, and leads to instantaneous improvement of extension lag, without worsening knee joint pain. HAL-SJ-based knee extension training could represent a viable novel post-TKA rehabilitation modality.