Study design and setting
The RWD database, which is administered by the Health, Clinic and Education Information Evaluation Institute (HCEI, Kyoto, Japan), was used for this retrospective cohort study [12, 13]. The HCEI is a not-for-profit research service foundation in Japan. Real World Data Co., Ltd. (Kyoto, Japan) supports the HCEI for data collection and standardization support. The RWD database includes the data of 20 million patients from 160 hospitals in Japan. This information consists of demographic data, prescriptions, laboratory results, diagnoses with International Statistical Classification of Diseases and Related Health Problems, 10th Revision Codes (ICD-10 codes) and inpatient and outpatient procedures. These data are handled by allocating a unique identifier to each individual. We did not link these data with any other databases.
Data Collection And Definitions
We included inpatients aged 18 or older who were prescribed RAASis for at least one month prior to admission and were undergoing CAG during hospitalization between April 2005 and March 2019. Only first admissions with CAG were included in our study. To restrict the focus to patients at risk of AKI, only patients with an estimated glomerular filtration rate (eGFR) of 15-60 ml/min/1.73 m2 were included. Patients undergoing dialysis within 3 days after admission were excluded. Patients with missing values for the primary diagnosis were also excluded because the primary diagnosis was treated as one of the explanatory variables in a previous study and may influence the incidence of AKI . The exposure of interest was the presence of a dose reduction in RAASis during the 3 days before CAG was performed. The definition of dose reduction in RAASis was the change in the ratio of the prescribed dose to the defined daily dose (dose/DDD), which is defined by the World Health Organization [15, 16]. We calculated the dose/DDD of RAASis for all days during hospitalization. If the dose/DDD was reduced, the patient was considered to have received a dose reduction of the RAASi and was included in the reduction group. For example, even if there was a change from angiotensin receptor blocker (ARB) to angiotensin-converting enzyme inhibitor (ACEi), we considered that as no reduction unless there was a change in the dose/DDD. We included both discontinuation and dose reduction in the exposure group to avoid misclassification of discontinuation of only one medication for patients prescribed two or more RAASis. The difference in the impact on the development of AKI between discontinuation and reduction (excluding discontinuation) was assessed in the subgroup analysis.
Propensity Score Matching
We calculated the propensity score (PS) of each patients to balance the baseline characteristics of each group [17–19]. The PS was estimated using a logistic regression model. The explanatory variables for the model are age, sex, Charlson Comorbidity Index (CCI) score, intensive care unit admission, baseline serum creatinine defined as the last measured serum creatinine before CAG, primary diagnosis on admission , infection , acute heart failure, hyponatremia, transfusion, hydration with extracellular fluid, N-acetyl cysteine, platinum-based chemotherapy, and the use of diuretics, amphotericin B, aminoglycosides, glycopeptides, and nonsteroidal anti-inflammatory drugs. The definitions of these explanatory variables are described in Supplementary data, Table S1. To ensure balancing between the two groups, calipers were defined as 0.1, and 1:2 nearest neighbor matching was applied to maintain statistical power. Standardized differences were used to assess comparability between the two matched cohorts.
Sample Size Calculation
We also estimated the needed sample size to detect the difference in the development of AKI between the two groups because the present study was planned based on the hypothesis that previous studies did not include a sufficient number of patients. The incidence of AKI was set at 10% in the reduction group and 15% in the control group [1–3]. The effect size was based on a systematic review of 3 RCTs . The required sample size was calculated to be 525 in the reduction group and 1,050 in the control group, assuming an alpha error of 0.05, a power of 0.8, and an enrollment ratio of 1:2.
The primary outcome was AKI defined as an absolute increase in serum creatinine of ≥ 0.3 mg/dl from baseline within 48 hours or a relative increase in serum creatinine of ≥ 50% within 7 days . We adopted the definition of AKI as renal injury after contrast use. This is because the Kidney Disease: Improving Global Outcome (KDIGO) guidelines indicated that contrast-induced nephropathy should be evaluated under the same criteria as AKI . The secondary outcomes were the need for dialysis and in-hospital mortality.
Continuous variables were reported as the mean and standard deviation, and categorical variables were reported as numbers and percentages. First, we described the characteristics of the patients in each group. Binary logistic regression was used to assess significant univariate associations between the reduction group and the control group with the outcomes for the PS-matched cohort. The results are presented as odds ratios (ORs) and 95% confidence intervals (CIs). We defined the main result as the OR in the PS-matched cohort.
Sensitivity analysis was performed by changing the period of time of dose reduction from 1 to 14 days to detect the appropriate timing of dose reduction of RAASis. We also performed subgroup analysis investigating the interaction with age (≤ 65 and > 65), type of RAASi (ARB, ACEi, aldosterone blocker and direct renin inhibitor (DRI)), past history of diabetes, baseline eGFR (≤ 30, 30-45 and > 45 ml/min/1.73 m2) and dose reduction or discontinuation. P-values with a two-sided test were reported, and P <0.05 was considered to indicate statistical significance. We used R ver. 4.1.2 to perform all statistical analyses.