Objective: We aimed to evaluate the important potential risks listed in the risk management plans of SGLT–2 inhibitors (SGLT–2i) using a real world database.

Design: A cohort study of patients prescribed either SGLT–2i (exposure group) or DPP–4 inhibitors (DPP–4i, control group), using a large–scale health insurance database including claims and specific health checkup.

Methods: The study population included the patients with type 2 diabetes between April 2014 and August 2021, and received either SGLT–2i or DPP–4i monotherapy, based on the claims in the database provided by DeSC Healthcare, Inc．The comparability between treatment groups was ensured by propensity score matching (PSM) and inverse probability treatment weighting (IPTW)．The outcome events included liver disorder, malignant tumors, fractures, cardiovascular disease, acute pancreatitis, acute kidney injury, and lower limb amputation. Hazard ratios (HRs) were estimated using the Cox proportional hazards model, in addition to five types of bias analyses.

Results: In the PSM population, the HRs (95% confidential interval [CI]) of the SGLT–2i group versus the DPP–4i group were 0.63 (0.28–1.44) for acute kidney injury, 0.75 (0.58–0.95) for fractures, 0.85 (0.67–1.07) for liver disorders, 0.87 (0.71–1.05) for cardiovascular diseases, 1.15 (0.88–1.51) for malignant tumors, 1.51 (0.71–3.19) for acute pancreatitis, and with no observation of lower limb amputation．In the IPTW population, the HRs (95% CI) of the SGLT–2i group versus the DPP–4i group were 0.72 (0.47–1.10) for acute kidney injury, 0.76 (0.67–0.86) for fractures, 0.91 (0.80–1.02) for liver disorders, 0.86 (0.78–0.94) for cardiovascular diseases, 1.04 (0.92–1.18) for malignant tumors, 1.81 (1.24–2.64) for acute pancreatitis, and 2.89 (0.69–12.1) for lower limb amputation. The ad hoc analysis of malignant tumors by type revealed several organ–specific statistically significant increases or decreases in HRs among the IPTW subjects; however, no significant overall HR for malignant tumors was observed. Some of the bias analysis revealed that there were significant decreases in HRs for acute kidney injury and liver disorders and a significant increase in HR for lower limb amputation.

Conclusion: In comparison to DPP–4i, the use of SGLT–2i was not associated with overall risk of malignant tumors. Further confirmatory studies with fit–for–purpose design are warranted to verify the potential, increased or decreased organ–specific risks of malignant tumors by type and the results suggesting decreased risks of bone fracture and cardiovascular diseases, and an increased risk of acute pancreatitis. (Jpn J Pharmacoepidemiol 2026; 31(1): 1‒18)

Introduction: The first–line agent for the treatment of ulcerative colitis is 5–ASA (5–aminosalicylic acid). About 10% of patients taking 5–ASA are deemed to develop a condition called “5–ASA intolerance,” in which they have difficulty taking 5–ASA continuously due to adverse effects. The incidence of 5–ASA intolerance seems to be on the rise. We can provide safer treatment if we can identify patients at high risk of developing 5–ASA intolerance, but there are few prediction models for the 5–ASA intolerance.

Objective: The purpose of this study was to develop and internally validate a prediction model for the 5–ASA intolerance among Japanese ulcerative colitis patients using real–world data.

Methods: We analyzed data from January 2005 to March 2023 using the payer database held by JMDC Inc. Japanese patients aged 15 years and older who were diagnosed with ulcerative colitis and prescribed oral 5–ASA were included in the analysis. Index date was defined as the date 5–ASA was first dispensed. A prediction model was developed using a Cox proportional hazards model with the number of days from the index date to the occurrence of an event (5–ASA intolerance) as the outcome. Predictors were selected based on expert opinions and the results of Cox regression. Internal validity of the model was assessed from two points; 1) The model’s discriminative ability by optimism–corrected c–index with bootstrapping, 2) The model’s accuracy by a calibration plot.

Results: The sample size was 9,520 with 931 events. The selected predictors were gender, age, oral 5–ASA brands, oral 5–ASA prescription dose, and the presence of a diagnosis of a certain disease (i.e. intestinal infection, influenza or pneumonia, iron–deficiency anemia, gastro esophageal reflux disease, gastric ulcer, and acute pancreatitis) within the past 1 year of the Index date. The optimism–corrected c–index was 0.5934. The calibration plot shows adequate fit.

Conclusion: With the developed prediction model, we could identify patients with ulcerative colitis who are at high risk for 5–ASA intolerance.: (Jpn J Pharmacoepidemiol 2026; 31 (1): 19‒29)

Immortal time in observational studies arises when three core design components—eligibility determination (T_elig), treatment assignment (T_assign), and the start of follow-up (time zero, T₀)—are not temporally aligned, which can substantially distort estimates of treatment effects. Based on the target trial framework, this paper provides a structured overview of design options to avoid immortal time, centered on the principle of synchronizing these three elements at T₀. We classify common observational scenarios into four types along two dimensions: (i) how often eligibility is met and (ii) whether treatment strategies can be distinguished at T₀. We then discuss sequential trial emulation, which can improve statistical efficiency when eligibility is repeatedly satisfied, and the clone-censor-weight method, which can accommodate complex treatment strategies that include a grace period, illustrating each with applied examples. We also situate the new-user design and its variants, as well as conventional approaches to immortal time such as landmark and person-time analyses, by clarifying their theoretical connections to and departures from target trial emulation. The contribution of the approaches presented here extends beyond the technical avoidance of immortal time. Their central value lies in enabling researchers to formulate causal questions while preserving clinically meaningful treatment strategies that explicitly specify when treatment is initiated and how it is implemented in practice, rather than reshaping those questions to fit the constraints of traditional analytic methods. We hope that the systematic framework presented here will catalyze a paradigm shift in applied observational research and serve as a standard for evidence generation that supports clinical decision-making. : (Jpn J Pharmacoepidemiol 2026; 31 (1): 31‒45)