Sodium-Glucose Cotransport Protein 2 Inhibitors in Patients With Type 2 Diabetes and Acute Kidney Disease

Key Points Question Do sodium-glucose cotransport protein 2 inhibitors (SGLT-2is) have beneficial associations with mortality, major kidney events (MAKEs), and major adverse cardiovascular events (MACEs) in patients with type 2 diabetes and acute kidney disease (AKD)? Findings In this cohort study of 230 366 patients, SGLT-2i use among those with type 2 diabetes and AKD was associated with significantly lower risks of mortality, MAKEs, and MACEs compared with nonuse. Meaning These findings suggest that the use of SGLT-2is in patients with type 2 diabetes and AKD is associated with reduced mortality, MAKEs, and MACEs, highlighting their potential clinical implications.

drug.Medications may also be coded to national drug codes or to anatomic therapeutic chemical codes, used in many European countries, or local codes.Additionally, laboratory information systems at HCOs and commercial laboratories rarely use standard codes, such as Logical Observation Identifiers Names and Codes (LOINCs), for test results.

DATA QUALITY
Data quality is a major challenge when it comes to the proper use of research data, and can potentially compromise the validity of research results.(3) Although the adoption of Electronic Health Records (EHRs) has grown exponentially due to federal incentives and meaningful use requirements, the quality of the data within them, and therefore the data used for research, is still improving.As EHRs are primarily designed for billing and patient care functions, the data may not be of the highest quality for research purposes.16To address this issue, a comprehensive data-quality framework and approach are needed.(4) Research on data quality is limited and generally focuses on assessing the quality of data in a single system or institution, (5)as well as determining if the data is good enough for its primary purpose, which is providing clinical care to patients.TriNetX has developed a comprehensive methodology to assess the quality of the data it uses.This methodology, which consists of four Cs-cleanliness, consistency, correctness, and completeness-takes into account the data extracted from the source systems, which is then transformed, cleaned-up, deduplicated, de-identified, optionally obfuscated, and semantically mapped.

ANALYSIS SPECIFICATIONS
The Compare Outcomes Analytic offers four different types of analyses: Measure of Association, Survival, Number of Instances, and Lab Result Distribution.The first three analyses have the option to "exclude patients with outcomes prior to the window."This option is useful when analyzing outcomes that are chronic diseases, as patients who have already developed the outcome are not at risk of developing it during the time window.When the "exclude patients with outcomes prior to the time window" option is unchecked, all patients in the cohort will be included in the analysis, regardless of whether they had the outcome prior to the time window.However, if this option is checked, patients will be excluded if their medical record indicates that they had the outcome before the start of the time window.This exclusion will apply to all patients who had the outcome prior to the index event, and any patients who develop the outcome between the index event and the start of the time window will also be excluded if the time window starts some days after the index event.

MEASURE OF ASSOCIATION ANALYSIS
The Measure of Association Analysis assesses the fraction of patients with a specified outcome.
The output summary includes the number of patients in each cohort meeting the query criteria, the number of patients with the outcome in each cohort, and the risk of the outcome in each cohort.
We made adjustments for various factors including demographics (age, sex, race), health conditions (underlying chronic kidney disease, hyperuricemia, congestive heart failure, ischemic heart diseases, cerebrovascular diseases, overweight, chronic obstructive pulmonary disease, musculoskeletal disease, malignancy), medications (Metformin, Sulfonylureas, Acarbose, Insulin, Aspirin, Clopidogrel, Atrovastatin, Allopurinol, Febuxostat, alpha-blocker, beta-blocker, calcium channel blocker), and clinical metrics (body mass index, leukocyte count, platelet count, estimated glomerular rate, proteinuria level, total cholesterol, glycohemoglobin, aspartate transaminase, B-type natriuretic peptide).Furthermore, the output includes the Risk Difference (the difference in the risks in cohorts of type 2 diabetes with AKD who received and not received SGLT-2 inhibitors), Risk Ratio (the ratio of the risks in SGLT-2i users cohort and non-user cohort), and Odds Ratio (the ratio of the odds in SGLT-2i users cohort and non-user cohort).

SURVIVAL ANALYSIS
The Kaplan-Meier Analysis estimates the probability of the outcome at a respective time interval (in this analysis, daily time intervals are used).To account for patients who exited the cohort during the analysis period, censoring is applied, whereby these patients are removed from the analysis after the last fact in their record.The output summary includes the number of patients in each cohort (meeting the query criteria), the number of patients with the outcome in the time window, median survival (the number of days when survival drops below 50%; the "-" indicates survival does not drop below 50% during the time window), and survival probability at the end of the window (the % survival at the end of the window).Furthermore, the Log-Rank Test, Hazard Ratio, and test for Proportionality are employed.

NUMBER OF INSTANCES ANALYSIS
The Number of Instances Analysis is a method that calculates the frequency of an outcome within a given time window.This analysis includes two settings: patients with zero instances and the definition of an instance.When selecting to exclude patients with zero instances, these patients are not included in the calculations for mean number of instances, standard deviation, or median.The histogram that displays the distribution of patients by number of instances will not include a bar for zero.Alternatively, by selecting to include patients with zero instances, the mean, standard deviation, and median for the number of instances will reflect the entire patient population, including those with zero instances.The histogram will include a bar for zero instances.The definition of an instance impacts how the counts are analyzed.When selecting Date, each calendar date on which any of the terms selected in the outcome are recorded will represent one instance.For instance, if the outcome is "Med A or Med B," and a patient has "Med A" on January 3, both medications on January 4, and "Med B" on January 6, then that patient is considered to have three instances, representing January 3, January 4, and January 6.It is important to note that if an outcome occurs across multiple dates, only the start date is tracked for the purpose of counting instances.For instance, a patient who begins a hospital stay on January 1, ends on January 3, begins another stay on January 10, and ends on January 15, is considered to have two instances of the outcome.Selecting Visit as an instance will count any visit that includes the outcome as one instance, regardless of how many times it occurred.
For example, a patient administered an analgesic on each of the three days of an inpatient stay following some index event.If analgesic is an outcome, these three administrations will represent only one instance because they are associated with the same visit.The output summary includes the count of patients in the cohort, the count of patients in the cohort that had the outcome in the time window, the mean, standard deviation, and median of the counts, and the median (1+ instances) when patients with zero instances are included in the analysis.Additionally, T-Test statistics testing for the difference between the cohorts are included.

LABORATORY RESULTS ANALYSIS
In the analysis, only lab results that are relevant to the outcomes are included.Furthermore, only the most recent lab values within the time window are taken into account.For numeric lab results, the outcome summary contains the number of patients in the cohort who meet the query criteria, the number of patients with the outcome within the time window, the mean, and the standard deviation of the lab values in the cohort.Additionally, T-Test statistics are provided to evaluate the difference between the cohorts.For non-numeric lab results, the counts of Negatives, Positives, and Unknowns are reported, and these percentages of the total counts are represented in the form of a bar chart.

LIMITATION
The TriNetX platform operates on individual-level data.However, as researchers, we do not have direct access to this individual data.Instead, the platform provides us with aggregated counts and statistical summaries of de-identified data, ensuring the privacy and confidentiality of the individuals represented in the data.The TriNetX platform pools data from various participating institutions, and while it processes individual-level data, it only allows researchers like us to interact with and analyze the aggregated and summarized data.This ensures compliance with both the Health Insurance Portability and Accountability Act and the General Data Protection Regulation.

External validation
We utilized data from the Chang Gung Research Database (CGRD), which is the most extensive collection of electronic medical records (EMR) across multiple institutions in Taiwan (6).This database provides researchers with convenient access to standardized patient-level data, enabling efficient utilization for a range of studies related to SGLT2 inhibitors.These studies encompass investigations into renoprotective and cardioprotective of SGLT2 inhibitors, as well as the assessment of their adverse side effects (7,8,9,10,11).The CGRD is a comprehensive collection of daily medical records gathered prospectively from seven branches of Chang Gung Memorial Hospital in Taiwan since January 2001.This database encompasses a significant volume of medical information, with an average of 500,000 emergency department visits, 8,500,000 outpatient visits, and over 280,000 admissions to 10,070 beds annually (6).The CGRD includes detailed personal information about patients, such as gender, body weight, height, lifestyle, and birth date.It also contains laboratory findings, pathology reports, imaging exam results, and comprehensive information about every emergency, inpatient, and outpatient visit.The database uses International Classification of Diseases, 9th and 10th revision, Clinical Modification (ICD-9-CM and ICD-10-CM) codes for classifying underlying diseases, reasons for admission, and details of emergency and outpatient visits.To prioritize patient privacy, the chart number of each patient was encrypted and exclusively utilized for data linkage between different databases within the CGRD (12).Our study received approval from the Institutional Review Board of the Chang Gung Medical Foundation (IRB No.: 201702274B0), ensuring adherence to ethical guidelines and patient confidentiality.The CGRD's extensive collection of medical information made it an optimal resource for conducting retrospective clinical studies, greatly facilitating our research efforts.

Sensitivity, specificity, positive outcome controls, and negative outcome controls
To ensure the reliability of our findings, several sensitivity analyses were carried out.Firstly, we investigated the relationship between variables across different enrollment periods to account for possible changes in antihyperglycemic prescribing preferences over time.Additionally, we excluded patients with short follow-up durations and those who experienced mortality at different times during the follow-up period.Our study utilized global healthcare data from TriNetX, spanning from September 2002 to September 2022.Notably, between 2012 and 2015, the European Drug Administration (EMA) and the US Food and Drug Administration (FDA) approved three SGLT2 inhibitors (Kangliflozin, Daggliflozin, and Enggliflozin) for lowering blood glucose levels in type 2 diabetes patients.Considering the potential selection bias for patients in the non-SGLT2 inhibitors group before 2012, we conducted a distinct sensitivity analysis to address this concern.
Moreover, we incorporated subjects who experienced events within 90 days post-discharge during the AKD phases in another sensitivity analysis.Secondly, diverse covariates were included in the Cox regression models to further validate the robustness of our results within each cohort.Furthermore, we performed specificity analyses to examine the beneficial effect of SGLT-2 inhibitors on different composite adverse outcomes.Thirdly, we utilized a new-user design focusing on those newly initiating SGLT-2i and juxtaposed this with individuals newly starting other active treatments, namely Sulfonylureas, dipeptidyl peptidase-4 inhibitors, or Pioglitazone.
To evaluate the effectiveness of our approach in reproducing known associations, we tested diabetes ketoacidosis and osteoporotic fractures as positive outcome controls.Additionally, we explored the correlation between SGLT-2 inhibitors treatment and five unrelated events (atopic dermatitis, conjunctivitis, melanoma, lymphoma, and Hodgkin's disease) as specified negative outcome controls.No prior evidence suggested a causal relationship between SGLT-2 inhibitors and the specified negative outcome controls.Sensitivity, specificity analysis, positive outcome controls and negative outcome controls were performed using R software (version 3.

©2024
Pan HC et al.JAMA Network Open.

©2024
Pan HC et al.JAMA Network Open.

1 .
2.2, Free Software Foundation, Inc, Boston, MA), SAS software (version 9.2, SAS Inc., Cary, NC), and Stata/MP software (version 16, StataCorp, College Station, TX) s part of our rigorous analytical approach.(A) Cohorts of type 2 diabetes with AKD who received and not received SGLT-2 inhibitors Cohort of type 2 diabetes with AKD who received SGLT-2 inhibitors This query was run on the network Research with 76 HCO(s) queried and 76 HCO(s) responded.

Cohort of type 2 diabetes with AKD who did not receive SGLT-2 inhibitors
©2024 Pan HC et al.JAMA Network Open.©2024Pan HC et al.JAMA Network Open.©2024Pan HC et al.JAMA Network Open.2.©2024 Pan HC et al.JAMA Network Open.©2024 Pan HC et al.JAMA Network Open.©2024 Pan HC et al.JAMA Network Open.©2024 Pan HC et al.JAMA Network Open.

Kidney Function and Electrolytes After Withdrawal of Dialysis
Risk of Mortality in Patients With Type 2 Diabetes and AKD: Comparison Between SGLT-2I Users and Nonusers After Propensity Score Matching The table presents results from a cohort analysis after propensity score matching.Cohort 1 consist of type 2 diabetes patients with AKD who received SGLT-2 inhibitors, while cohort 2 comprises those who did not receive of SGLT-2 inhibitors.Risk of MAKE in Patients With Type 2 Diabetes and AKD: Comparison Between SGLT-2I Users and Nonusers After Propensity Score Matching The table presents results from a cohort analysis after propensity score matching.Patients with a history of MAKE were part of the baseline cohort but were excluded from this analysis.Cohort 1 Comparing SGLT-2I Users and Nonusers in Relation to MACE Risk of MACE in Patients With Type 2 Diabetes and AKD: Comparison Between SGLT-2I Users and Nonusers After Propensity Score MatchingThe table presents results from a cohort analysis after propensity score matching.Patients with a patients in Cohort 1 and 2,827 patients in Cohort 2 were excluded from results because they had the outcome prior to the time window.

Meier survival analysis excluding patients with outcome prior to the time window
,765 patients in Cohort 1 and 2,827 patients in Cohort 2 were excluded from results because they had the outcome prior to the time window.