This retrospective, multicentric, observational cohort study using common data model was supported by Observational Health Data Sciences and Informatics (OHDSI) collaborative. The OHDSI has been providing database software based on Observational Medical Outcomes Partnership (OMOP) CDM17. We used data from three referral hospitals’ electronic medical records (EMR) databases: Seoul National University Bundang Hospital (SNUBH), Asan Medical center (AMC), and Seoul National University Hospital (SNUH). All three participated institutions are large medical centers located in urban area of South Korea. The EMR of SNUBH, the oldest database, consisted of data from 2003 to 2019. Other two databases (AMC and SNUH) had complete EMR data from 2004 to 2019. All EMR data from each hospital were transformed into OMOP-CDM version 5 schema18. The OMOP-CDM database comprised automatically deidentified EMR data including prescriptions, laboratory test, diagnoses, procedures, and demographics of visiting patients. The CDM schema enables identical data analysis execution on heterogenous EMR database using transformed standardized vocabulary.
Study design and cohort construction
This retrospective, multicenter, observational study employed a comparative cohort design to assess effects of metformin on different groups based on various stages of renal function status. We identified patients with T2DM in each site using OMOP-CDM diagnosis codes. We then cross-validated them through HbA1c level ≥ 6.5% or fasting blood glucose ≥ 126 mg/dL to prevent over aggregation of study subjects. We defined index date (start of observation period) as the visit date of the first diagnosis of T2DM and end date of observation as the last time the study subject visited the hospital. We excluded patients who met the following condition: 1) a diagnosis of type 1 diabetes, 2) an age of less than 18 or over 80 at the index date, 3) prior exposure to corticosteroids for more than 12 weeks or calcineurin inhibitors, which could cause drug-induced diabetes, 4) a history of malignancies, ESKD, myocardial infarction, or cerebral infarction, and 5) had not prescribed any antidiabetic agent six months before or after the index date.
We identified cohorts as follows. First, T2DM cohorts were identified using a specific combination of laboratory tests and diagnosis code to identify diabetic patients who had not yet developed DN at the index date. Second, T2DM with each CKD stage (3A, 3B, and 4) cohort was identified by a combination of prior T2DM cohort definitions and renal function test using eGFR by creatinine-based EPI-CKD equation19.
To define the control cohort for investigating the effect of metformin on DN in patients newly diagnosed with T2DM, we applied the following approach within the common data model1. First, the index date was identified as the first hospital visit date which a diagnosis of T2DM was confirmed via diagnosis codes of ICD-10 and HbA1c level. Second, considering standard clinical practice, patients prescribed metformin in the 180 days surrounding the index date (both prior and subsequent) who continued metformin treatment at least once beyond the initial 180 days were designated as the metformin-using group. Controls were patients who were prescribed other OHAs within the 180-day window surrounding the index date without receiving any further metformin prescriptions after this 180-day period throughout the observation period. Visual explanations of the study cohorts are shown in Fig. 1(a).
To analyze outcomes in T2DM patients with eGFR levels less than 60, 45, and 30 ml/min/1.73 m2, respectively, control cohorts were defined in a similar manner. For a patient diagnosed with T2DM (as determined by a specific diagnosis code and HbA1c level) who had two or more confirmed readings of eGFR less than 60, 45, or 30, the index date was established as the time of the first eGFR measurement that fell below the respective threshold. Considering conventional clinical practice for DN patients showing diminished renal function, the metformin user group was defined as those who were prescribed metformin in the 90-day period both prior to and after the index date and continued the use of metformin at least once beyond this 90-day period. The control group consisted of patients who were prescribed OHAs within the 90-day timeframe surrounding the index date without receiving any further metformin prescriptions after this 90-day period during the observation period. Visual explanations of study cohorts are shown in Fig. 1b. After identifying cohorts, four sets of treatment–control comparative cohorts were established within each participating institution. This stratification was designed to elucidate influence of metformin on diverse groups categorized according to respective renal function status.
Outcomes
Primary outcomes of this study were net major adverse cardiovascular events (MACE) or in-hospital death and a composite of major adverse kidney events (MAKE) or in-hospital death. MACEs were defined as ischemic events including myocardial infarction and ischemic stroke as determined by diagnostic history. MAKE were defined as the need for renal replacement therapy (RRT) or an eGFR of less than 15 ml/min/1.73 m2 for more than two consecutive measurements based on the EPI-CKD equation. In addition, this study evaluated the incidence of new-onset DN or overt DN in the T2DM without DN cohort. We defined DN as either spot urine albumin-to-creatinine ratio of 30 mg/g or 24-h albuminuria of 30 mg or more, either spot urine protein-to-creatinine ratio of 150 mg/g or 24-h proteinuria of 150 mg or more, or EPI-CKD eGFR less than 60 ml/min/1.73 m2. Overt DN definition was either spot urine albumin-to-creatinine ratio of 300 mg/g or 24-h albuminuria of 300 mg or more, either spot urine protein-to-creatinine ratio of 500 mg/g or 24-h proteinuria of 500 mg or more, or EPI-CKD eGFR less than 60 ml/min/1.73 m2.
Statistical analysis
We used a 1:1 matching of propensity score (PS) to reduce selection bias due to differences between treatment and control groups. PS was estimated by L1 logistic regression adjusted by tenfold cross validation with a caliper of 0.2 on the logit scale. The Cyclops R package (https://github.com/ohdsi/cyclops) was used for PS matching. We used sex, age groups, drug prescriptions other than the target drug, metformin, and disease diagnosis before the index date as covariates for the matching. Measurements and conditions after the index date were not used for matching. We used the Cox proportional hazard model to analyze matched cohorts and calculate hazard ratios (HR) along with 95% confidence intervals (CI) for outcomes. Survival curves were constructed and compared using Kaplan–Meier estimates and log-rank tests. We then conducted meta-analysis to calculate pooled estimated HRs across the three databases. Meta-analysis was conducted using a random-effect model with heterogeneity analysis with I2 index20. To assess robustness, components of outcome such as sensitivity were analyzed21. All statistical tests were performed using R (version 3.6.3; the R foundation for Statistical Computing), including packages provided by the OHDSI collaboration22.
Ethical approval
This study was approved by the ethics committee of each institution, and the requirement for informed consent was waived by the Institutional Review Board of Seoul National University Bundang Hospital (IRB No. X-1808–484-906). This study met the criteria for waiver of informed consent as stated in the Bioethics and Safety Act of the Republic of Korea, and all methods used were performed according to the relevant guidelines and regulations.
Ethics approval and consent to participate
The study was approved by the research ethics committees of the participating centers: The Institutional Review Board of Seoul National University Bundang Hospital, Asan Medical center, and Seoul National University.
- The Renal Warrior Project. Join Now
- Source: https://www.nature.com/articles/s41598-024-52078-4