Abstract
Background
Although sodium-glucose cotransporter 2 inhibitors (SGLT2is) slow the progression of kidney disease in diabetic patients, whether certain subsets of diabetic patients present heterogeneous phenotypes, such as a rapid deterioration of kidney function, following SGLT2i use remains unclear. This study aimed to clarify these heterogeneous responses and identify the factors contributing to them through a kidney function trajectory analysis.
Methods
This retrospective cohort study included 66,375 patients who visited the diabetes clinic for type 2 diabetes management, 5,209 of whom were prescribed SGLT2is. The median duration of SGLT2i prescription was 29.0 months (interquartile range, 16.3 to 49.0 months). Patterns of estimated glomerular filtration rate (eGFR) trajectories were analyzed using a latent class linear mixed model for 4,185 patients, with a median follow-up duration of 29.7 months (interquartile range, 17.1 to 34.1 months) and factors associated with these patterns were evaluated via multinomial logistic regression analysis.
Results
Four distinct eGFR trajectory patterns were identified: 3,636 (86.9%) in class 1 (high baseline, stable eGFR), 440 (10.5%) in class 2 (low baseline, stable eGFR), 73 (1.7%) in class 3 (high baseline, rapidly declining eGFR), and 36 (0.9%) in class 4 (low baseline, rapidly declining eGFR). A rapid eGFR decline was more common in younger patients with substantial albuminuria, and hypoalbuminemia. Patients in the rapidly declining eGFR classes had higher risks of hospitalization and death than those in stable eGFR classes did.
Conclusion
The heterogeneity in eGFR trajectories following SGLT2i use suggests that not all patients benefit equally from this therapy. Identifying patients at risk of a rapid eGFR decline is critical for providing timely warnings and optimizing treatment strategies.
Keywords: Diabetes Mellitus, Diabetes Complications, Diabetic Nephropathies, Sodium-Glucose Transporter 2 Inhibitors, Estimated Glomerular Filtration Rate (eGFR)
Graphical Abstract
INTRODUCTION
Diabetes affects 10% of individuals aged 20–79 years (over 500 million people) worldwide, contributing to a substantial increase in the incidence of diabetic kidney disease (DKD) and end-stage kidney disease (ESKD), with approximately 50% of ESKD cases attributable to diabetes.1,2 The increasing global prevalence of these conditions is associated with substantial risks of all-cause and cardiovascular mortality.1,3,4 Managing these diseases and improving patient outcomes remain critical challenges in health care. As progression to ESKD is a severe consequence for diabetic patients, understanding and mitigating the risks associated with DKD is essential for improving future prognoses.
Sodium-glucose cotransporter 2 inhibitors (SGLT2is), such as empagliflozin, dapagliflozin, canagliflozin, ertugliflozin, and bexagliflozin, have emerged as essential therapeutic agents in the management of diabetes and its cardiorenal complications.5 Clinical trials have documented the efficacy of SGLT2is in slowing the progression of kidney disease and reducing the risk of cardiovascular events in diabetic patients.6,7 These studies highlight the potential of SGLT2is to alter the course of DKD to the point where patients are increasingly willing to use these medications as part of their treatment. As a result, the recent Kidney Disease Improving Global Outcomes guidelines recommend SGLT2is as a first-line treatment for diabetic patients, particularly those with an estimated glomerular filtration rate (eGFR) of at least 20 mL/min/1.73 m2.8
While DKD is typically diagnosed through the onset of persistent albuminuria followed by a gradual decrease in the eGFR, patients exhibit heterogeneous phenotypes with varying clinical presentations and disease patterns.9,10,11 This variability is reflected in different rates of kidney function decline and treatment responses. Some DKD patients experience a slow progression of kidney damage, maintaining stable kidney function for many years, whereas others rapidly progress to ESKD.10,12,13 Most patients present with predominant albuminuria and either stable or declining eGFR, whereas a small subset shows a rapid eGFR decline with minimal or no albuminuria.12,13 Among patients with type 2 diabetes, the median eGFR slope typically ranges from −1 to −4 mL/min/1.73 m2 per year.14,15 Rapid decliners, defined as those with an annual eGFR decline of ≥ 5 mL/min/1.73 m2, are particularly vulnerable.16 Factors contributing to this heterogeneity include genetic predispositions, comorbidities such as hypertension and cardiovascular disease, variability in glycemic control, and differences in the underlying mechanisms driving kidney damage.17,18,19 Although SGLT2i therapy is widely used in DKD management, individual responses to treatment may vary due to the above factors.20 The diverse patterns of DKD progression underscore the need for a more personalized treatment approach, emphasizing the importance of identifying patients at risk of rapid decline. This study aims to explore the heterogeneous progression of kidney disease in patients with type 2 diabetes following SGLT2i use by applying a trajectory model. By identifying the factors associated with these progression patterns and evaluating their relationships with hospitalization and death outcomes, we may provide insights that could inform personalized treatment strategies and improve the prognosis of diabetic patients.
METHODS
Study population
Data were collected from the clinical data warehouse of Seoul National University Hospital, a tertiary care center in South Korea. The study identified 66,375 individuals with diabetes who visited the diabetes clinic between January 2003 and October 2023. Among them, 8,254 patients were prescribed empagliflozin or dapagliflozin for the first time, the only two SGLT2is available in South Korea. The remaining 58,121 patients had not been prescribed SGLT2is. Patients aged less than 18 years old, those with type 1 or other types of diabetes, those receiving dialysis or kidney transplantation, those with an eGFR below 20 mL/min/1.73 m2, or those with less than one year of eGFR follow-up, were excluded. Additionally, for patients in the SGLT2i group, those prescribed SGLT2i for less than 30 days were excluded. After these exclusions, 5,209 patients remained in the SGLT2i group and 30,763 patients remained in the non-SGLT2i group. For the trajectory model, we excluded patients from the SGLT2i group who had only one average eGFR measurement after quarterly data merging (n = 255) and those without albuminuria follow-up during SGLT2i use (n = 769). Thus, 4,185 patients in the SGLT2i group were included in the trajectory analysis (Fig. 1).
Fig. 1. Flowchart for patient inclusion and exclusion.
SGLT2i = sodium-glucose cotransporter 2 inhibitor, ESKD = end-stage kidney disease, eGFR = estimated glomerular filtration rate, ACR = urine albumin-to-creatinine ratio.
Clinical variables
Clinical data, such as age, sex, eGFR, albuminuria, hemoglobin, serum albumin, hemoglobin A1c (HbA1c), total cholesterol, low-density lipoprotein (LDL) cholesterol, and a history of ischemic heart disease and stroke, were collected. The eGFR was calculated using the Chronic Kidney Disease Epidemiology Collaboration equation.21 Albuminuria was calculated as the urine albumin-to-creatinine ratio (ACR). If the urine protein-to-creatinine ratio (PCR) was available instead of the ACR, we converted the PCR to the ACR using a previously developed equation.22 Information on the use of various medications, such as glucose-lowering agents (e.g., metformin, sulfonylurea, thiazolidinedione, dipeptidyl peptidase-4 inhibitors, alpha-glucosidase inhibitors, insulin, and glucagon-like peptide 1 receptor agonists), antihypertensive agents (e.g., renin-angiotensin-aldosterone system inhibitors, diuretics, beta-blockers, and calcium channel blockers), statins, and antiplatelet agents was collected. The follow-up data were censored at the point of the last available measurement. The follow-up period was defined as the time from the initiation of SGLT2is in the SGLT2i group or from the baseline visit in the non-SGLT2i group to the date of the first study outcome or the censored date, whichever occurred first.
Statistical analysis
Patient information was analyzed using descriptive statistics, with the data presented as the means ± standard deviations or medians and interquartile ranges, as appropriate. Categorical variables are presented as numbers and percentages. Patients in the SGLT2i group were matched to those in the non-SGLT2i group using propensity scores to minimize the bias inherent in a retrospective cohort study. These scores were calculated via a multivariate logistic regression model. The covariates included in the propensity score model were age, sex, eGFR, albuminuria, hemoglobin, serum albumin, HbA1c, total cholesterol, LDL cholesterol, a history of ischemic heart disease and stroke, and the use of other medications. Matching between the SGLT2i and non-SGLT2i groups was performed via nearest 1:1 neighbor greedy matching. After propensity score matching, categorical variables were compared using the χ2 test or Fisher’s exact test, whereas continuous variables were compared using Student’s t-test or the Wilcoxon rank-sum test. We also calculated the standardized mean difference, which measures the difference between the means of two groups for a given variable. A standardized mean difference of 0.1 or greater typically indicates a meaningful imbalance between groups. The associations with study outcomes were evaluated using Kaplan-Meier curves and the log-rank test. Univariate and multivariate Cox proportional hazards models were used to determine hazard ratios (HRs) and confidence intervals (CIs) for the study outcomes. Age, sex, eGFR, and blood findings with P < 0.1 after propensity score matching were incorporated into the multivariate model. The kidney outcome was defined as a greater than 50% decrease in the eGFR from baseline or the development of ESKD. The composite outcome included either the kidney outcome or death.
Trajectory analysis
The analysis of eGFR trajectories was conducted by modeling quarterly average eGFRs using latent class linear mixed models (LCMMs). LCMMs extend the standard linear mixed model by accounting for expected heterogeneity in eGFR trajectories among patients. This method allows for flexibility, as it does not require the same number of observations for each patient or identical measurement time points. We analyzed eGFR data within 3 years after the initiation of SGLT2 inhibitors. eGFR measurements were taken until the discontinuation of SGLT2 inhibitors, development of ESKD, death, or the last available measurement, whichever occurred first. Repeated eGFR measurements were logarithmically transformed to approximate a Gaussian distribution. A random intercept term was specified in the model to account for individual deviations from the mean trajectory of each class. Covariates, including age, sex, SGLT2i dose, and albuminuria at each eGFR measurement time point during follow-up, were included to mitigate confounding effects. Missing albuminuria data were imputed using forward filling, which replaced missing data with the most recent available follow-up observations. If no initial albuminuria values were available, median filling based on the next available existing data was used.
We evaluated both linear and spline trajectory profiles. Models were built using a linear link function, I-splines with 7 nodes at quantiles, and I-splines with 7 equidistant nodes. Based on Akaike’s information criterion (AIC) and Bayesian information criterion (BIC), the I-spline with 7 nodes at quantiles was found to be superior, and this model was selected for further analysis. We fitted models with one to five classes and selected the optimal number of classes by considering the AIC, BIC, posterior probability of membership, sample size of each class, and clinical interpretability. After model convergence, each patient was assigned a posterior probability of belonging to each eGFR trajectory class. Patients were then classified into the group with the highest posterior likelihood of membership.
Factors associated with eGFR trajectories were evaluated using multivariate multinomial logistic regression with stepwise backward selection based on the AIC. The baseline covariates included sex, age, eGFR, albuminuria, hemoglobin, serum albumin, HbA1c, total cholesterol, LDL cholesterol, and a history of ischemic heart disease and stroke. The stable eGFR group was used as the reference group. We also assessed how changes in albuminuria differed across eGFR trajectory classes using a linear mixed model, with the trajectory class as a fixed effect and individual random slopes and intercepts. All analyses were conducted using R (version 4.4.0; R Foundation for Statistical Computing, Vienna, Austria) with the lcmm, survival, and lme4 packages.
Ethics statement
This single-center retrospective cohort study, conducted at Seoul National University Hospital, adhered to the principles of the Declaration of Helsinki. The study was approved by the Institutional Review Board of Seoul National University Hospital (No. H-2307-104-1449), and the requirement for informed consent was waived, as the analysis was based on anonymized datasets.
RESULTS
Patient characteristics
The baseline characteristics of the cohort are presented in Supplementary Table 1. The mean age of the 5,209 patients in the SGLT2i group was 59.6 ± 12.9 years, whereas it was 60.5 ± 11.6 years in the non-SGLT2i group. The median duration of SGLT2i prescription was 29.0 months (interquartile range, 16.3 to 49.0 months). Patients in the SGLT2i group presented higher baseline eGFRs and albuminuria levels than did those in the non-SGLT2i group. A history of ischemic heart disease was present in 20.2% of the study patients, with a higher prevalence in the SGLT2i group (30.9%) than in the non-SGLT2i group (18.4%).
After propensity score matching, the study included 3,818 patients in both the SGLT2i and non-SGLT2i groups, resulting in well-balanced baseline characteristics between the two groups (Supplementary Table 2). The mean age was similar in both groups, with a comparable proportion of male patients. No differences in the eGFR or albuminuria levels were observed between the groups. The HbA1c and total cholesterol levels were lower, whereas the LDL cholesterol level was higher in the SGLT2i group than in the non-SGLT2i group. A history of ischemic heart disease was more prevalent in the SGLT2i group than in the non-SGLT2i group, whereas the history of stroke was similar between the two groups.
Study outcomes according to SGLT2i use
The median duration of follow-up in the survival analysis was 47.6 months (interquartile range, 28.1 to 71.3 months; maximum, 116 months) in the SGLT2i group and 108.6 months (interquartile range, 57.8 to 172.3 months; maximum, 292 months) in the non-SGLT2i group. The Kaplan-Meier survival analysis revealed that the SGLT2i group presented better kidney and composite outcomes than the non-SGLT2i group did (P = 0.003) (Supplementary Fig. 1). According to the univariate Cox proportional hazard model, the SGLT2i group displayed a lower risk of kidney and composite outcomes than the non-SGLT2i group, with unadjusted HRs of 0.77 (95% CI, 0.64 to 0.92) for kidney outcomes (P = 0.003) and 0.75 (0.63 to 0.89) for composite outcomes (P = 0.001) (Supplementary Table 3). According to the multivariate model, the SGLT2i group consistently presented a lower risk of study outcomes than the non-SGLT2i group.
eGFR trajectories
The renoprotective effects of SGLT2is are well established, and these effects were observed in our cohort. We further identified the characteristics of a subset of patients who exhibited a diminished or heterogeneous response to SGLT2is. We conducted a trajectory analysis of the SGLT2i group, which included a total of 4,185 patients, to explore this finding. The mean age of the patients was 59.6 ± 12.9 years, and the median follow-up duration was 29.7 months (interquartile range, 17.1 to 34.1 months). Each patient had a median of 10 serum creatinine measurements (interquartile range, 6 to 13), totaling 57,635 measurements. After merging serial eGFR values into a single mean value for each three-month period, this number was reduced to an average of 8 serum creatinine measurements (interquartile range, 5 to 10), resulting in a total of 37,477 measurements used for trajectory modeling.
The characteristics of the models fitted with different numbers of trajectory classes (e.g., 2–5 classes) using linear and spline trajectory functions are presented in Supplementary Table 4. Four- and five-class spline trajectory models, which use I-splines with 7 nodes at quantiles, were selected as candidates because of their relatively lower BIC values (–111,617.6 for the four-class model and –111,380.4 for the five-class model). The four-class model was selected as all posterior probabilities exceeded 85%, and the smallest group accounted for 0.9%, making the model interpretable. Individual trajectories of eGFRs are shown in Supplementary Fig. 2.
A hypothetical individual with predefined characteristics, including age, sex, SGLT2i dose, and albuminuria, was assumed for the trajectory analysis. Fig. 2 illustrates the eGFR trajectories for four classes in a hypothetical 60-year-old man taking 10 mg of dapagliflozin with 300 mg/g albuminuria at baseline. According to the traditional linear mixed model, the rates of eGFR decline (in mL/min/1.73 m2 per year) were as follows: 0.18 (95% CI, −0.01 to 0.38) for class 1, 0.99 (0.44 to 1.55) for class 2, −13.51 (−14.37 to −12.65) for class 3, and −15.74 (−18.46 to −13.01) for class 4. Patients exhibited four distinct eGFR trajectories during the follow-up period, and the classes were named accordingly: class 1 (n = 3,636; 86.9%), characterized by a high baseline and stable eGFR; class 2 (n = 440; 10.5%), characterized by a low baseline and stable eGFR; and class 3 (n = 73; 1.7%) and class 4 (n = 36; 0.9%), characterized by rapid eGFR decline, with class 3 having a high baseline value and class 4 having a low baseline value.
Fig. 2. Trajectories of eGFRs and 95% condence intervals derived from a latent class linear mixed model. The model was adjusted for age, sex, doses of sodium-glucose cotransporter 2 inhibitors, and albuminuria at each measurement time point. A hypothetical model of a 60-year-old patient taking 10 mg of dapagliflozin with 300 mg/g albuminuria at baseline. The model classified patients into four classes: class 1 (high baseline and stable eGFR), class 2 (low baseline and stable eGFR), class 3 (high baseline and rapidly declining eGFR), and class 4 (low baseline and rapidly declining eGFR).
eGFR = estimated glomerular filtration rate.
Factors associated with eGFR trajectories
The baseline characteristics according to the adjusted eGFR trajectories are shown in Table 1. Compared with class 1 patients (high baseline and stable eGFR), class 3 patients (high baseline and rapidly declining eGFR) were young and had high levels of albuminuria and HbA1c and low levels of hemoglobin and serum albumin. Compared with class 2 patients (low baseline and stable eGFR), class 4 patients (low baseline and rapidly declining eGFR) were young; had a high baseline eGFR; high levels of albuminuria and cholesterol; and low levels of hemoglobin, serum albumin, and HbA1c.
Table 1. Baseline characteristics of the trajectory classes.
| Variables | Total (N = 4,185) | Class 1 (n = 3,636) | Class 2 (n = 440) | Class 3 (n = 73) | Class 4 (n = 36) | P value | |
|---|---|---|---|---|---|---|---|
| Age, yr | 59.6 ± 12.9 | 58.8 ± 12.6 | 60.0 ± 13.5 | 57.6 ± 14.1 | 58.3 ± 16.8 | < 0.208 | |
| Male | 2,619 (62.6) | 2,257 (62.1) | 296 (67.3) | 47 (64.4) | 19 (52.8) | 0.106 | |
| Blood findings | |||||||
| eGFR, mL/min/1.73 m2 | 83.4 ± 21.4 | 87.9 ± 17.5 | 49.3 ± 17.9 | 84.1 ± 25.4 | 58.5 ± 30.7 | < 0.001 | |
| ACR, mg/g | 224.7 ± 701.7 | 121.2 ± 379.7 | 768.18 ± 1,274.0 | 954.3 ± 1,760.1 | 2,834.0 ± 2,643.1 | < 0.001 | |
| Hemoglobin, g/dL | 14.1 ± 1.9 | 14.4 ± 1.7 | 13.1 ± 2.3 | 13.5 ± 2.2 | 12.2 ± 2.2 | < 0.001 | |
| Serum albumin, g/dL | 4.4 ± 0.4 | 4.43 ± 0.34 | 4.27 ± 0.42 | 4.07 ± 0.54 | 3.77 ± 0.74 | < 0.001 | |
| Hemoglobin A1c, % | 7.9 ± 1.3 | 7.9 ± 1.2 | 7.6 ± 1.4 | 8.1 ± 1.5 | 7.5 ± 1.6 | < 0.001 | |
| Total cholesterol, mg/dL | 158.3 ± 39.7 | 159.3 ± 39.5 | 159.6 ± 39.1 | 158.4 ± 40.2 | 172.1 ± 67.3 | 0.287 | |
| LDL cholesterol, mg/dL | 109.7 ± 40.3 | 111.2 ± 40.0 | 102.0 ± 41.5 | 97.4 ± 33.9 | 107.0 ± 50.0 | < 0.001 | |
| Comorbidities | |||||||
| Ischemic heart disease | 1,221 (29.2) | 1,069 (29.4) | 123 (28.0) | 20 (27.4) | 9 (25.0) | 0.844 | |
| Stroke | 746 (17.8) | 640 (17.6) | 80 (18.2) | 18 (24.7) | 8 (22.2) | 0.397 | |
| SGLT2 inhibitor | |||||||
| Dapagliflozin only | 2,161 (51.6) | 1,803 (49.6) | 290 (65.9) | 42 (57.5) | 26 (72.2) | < 0.001 | |
| Empagliflozin only | 1,955 (46.7) | 1,772 (48.7) | 144 (32.7) | 29 (39.7) | 10 (27.8) | < 0.001 | |
| Both | 69 (1.7) | 61 (1.7) | 6 (1.4) | 2 (2.7) | 0 (0) | 0.735 | |
| Other medication | |||||||
| Metformin | 3,054 (73.0) | 2,721 (74.8) | 251 (57.0) | 59 (80.8) | 23 (63.9) | < 0.001 | |
| Sulfonylurea | 2,396 (57.3) | 2,075 (57.1) | 255 (58.0) | 46 (63.0) | 20 (55.6) | 0.760 | |
| Thiazolidinedione | 321 (7.7) | 272 (7.5) | 40 (9.1) | 6 (8.2) | 3 (8.3) | 0.600 | |
| DPP-4 inhibitor | 1,334 (31.9) | 1,090 (30.0) | 187 (42.5) | 37 (50.7) | 20 (55.6) | < 0.001 | |
| Alpha-glucosidase inhibitor | 42 (1.0) | 32 (0.9) | 8 (1.8) | 1 (1.4) | 1 (2.8) | 0.094 | |
| Insulin | 1,663 (39.7) | 1,364 (37.5) | 220 (50.0) | 54 (74.0) | 25 (69.4) | < 0.001 | |
| GLP-1 receptor agonist | 535 (12.8) | 472 (13.0) | 51 (11.6) | 9 (12.3) | 3 (8.3) | 0.793 | |
| RAAS inhibitor | 2,538 (60.7) | 2,114 (58.1) | 342 (77.7) | 54 (74.0) | 28 (77.8) | < 0.001 | |
| Diuretics | 848 (20.3) | 626 (17.2) | 154 (35.0) | 42 (57.5) | 26 (72.2) | < 0.001 | |
| Beta blocker | 1,724 (41.2) | 1,441 (39.6) | 206 (46.8) | 47 (64.4) | 30 (83.3) | < 0.001 | |
| Calcium channel blocker | 1,608 (38.4) | 1,315 (36.2) | 224 (50.9) | 44 (60.3) | 25 (69.4) | < 0.001 | |
| Statin | 3,700 (88.4) | 3,223 (88.6) | 383 (87.0) | 63 (86.3) | 31 (86.1) | 0.590 | |
| Antiplatelet | 2,166 (51.8) | 1,880 (51.7) | 224 (50.9) | 43 (58.9) | 19 (52.8) | 0.651 | |
Values are presented as number (%) or mean ± standard deviation.
eGFR = estimated glomerular filtration rate, ACR = random urine albumin-to-creatinine ratio, LDL = low-density lipoprotein, SGLT2 = sodium-glucose cotransporter 2, DPP-4 = dipeptidyl peptidase-4, GLP-1 = glucagon-like peptide-1, RAAS = renin-angiotensin-aldosterone system.
Table 2 presents the associations between baseline factors and eGFR trajectories, as determined by multinomial logistic regression. Using backward stepwise variable selection adjusted for clinical covariates, the factors distinguishing eGFR patterns (e.g., class 1 vs. class 3 and class 2 vs. class 4) were baseline eGFR, albuminuria, and serum albumin. A young age was associated with a high odd of being in the declining eGFR group when comparing class 3 to class 1. Substantial albuminuria and low serum albumin levels were strong predictors of a rapid eGFR decline, with albuminuria (per 100 mg/g increase) consistently associated with this decline. When comparing class 4 to class 2, a higher baseline eGFR predicted a greater likelihood of belonging to the declining eGFR group.
Table 2. Factors associated with the pattern of eGFRs.
| Variables | Class 3 (vs. class 1) | Class 4 (vs. class 2) | ||
|---|---|---|---|---|
| OR (95% CI) | P value | OR (95% CI) | P value | |
| Age, per 1 yr | 0.95 (0.92–0.98) | 0.002 | ||
| Male (vs. female) | ||||
| eGFR, per 1 mL/min/1.73 m2 | 0.98 (0.96–1.00) | 0.025 | 1.08 (1.05–1.12) | < 0.001 |
| ACR, per 100 mg/g | 1.07 (1.05–1.10) | < 0.001 | 1.07 (1.04–1.10) | < 0.001 |
| Hemoglobin, per 1 mg/dL | ||||
| Serum albumin, per 1 mg/dL | 0.20 (0.11–0.37) | < 0.001 | 0.37 (0.16–0.87) | 0.023 |
| Hemoglobin A1c, per 1% | ||||
| Total cholesterol, per 1 mg/dL | ||||
| LDL cholesterol, per 1 mg/dL | ||||
| Ischemic heart disease (vs. none) | ||||
| Stroke (vs. none) | ||||
Backward stepwise variable selection was performed based on the Akaike Information Criteria. Classes 1 and 2 were used as the reference groups.
eGFR = estimated glomerular filtration rate, OR = odds ratio, CI = confidence interval, ACR = random urine albumin-to-creatinine ratio, LDL = low-density lipoprotein.
The albuminuria trajectories differed significantly among the trajectory classes (Fig. 3). Classes 1 and 2 had stable and low initial albuminuria. Class 3 patients started with low initial albuminuria, which then increased during the follow-up period. Class 4 patients presented high initial albuminuria, which continued to increase throughout the follow-up period. The characteristics of this albuminuria pattern changed in conjunction with the eGFR changes observed in each class.
Fig. 3. Trajectories of ACRs for the trajectory groups of estimated glomerular filtration rates.
ACR = random urine albumin-to-creatinine ratio.
Clinical outcomes associated with the trajectories
We compared clinical outcomes among the classes derived from the trajectory analysis and conducted survival analyses using both all-cause hospitalization alone and a composite outcome combining all-cause hospitalization and death. The occurrence of study outcomes was highest in class 3 and lowest in class 1 (Fig. 4). Compared with class 1, class 3 had an increased HR of 1.65 (95% CI, 1.41 to 1.92) for both all-cause hospitalization and the composite outcome in the unadjusted model. Class 4 had a 1.63 times increased HR of all-cause hospitalization (1.42 to 1.86) and a 1.62 times increased HR of composite outcome (1.42 to 1.85). After covariate adjustment, all of classes 2, 3, and 4 were associated with increased risks of all-cause hospitalization and composite outcome (Table 3).
Fig. 4. Kaplan-Meier survival plots for (A) all-cause hospitalization and (B) the composite outcome of hospitalization and death, stratified by trajectory groups of estimated glomerular filtration rates.
Table 3. Adjusted risk of outcomes according to the trajectory classes.
| Class | All-cause hospitalization | Composite outcome | ||||||
|---|---|---|---|---|---|---|---|---|
| No. of events | Events per 100 person-year | Hazard ratio (95% CI) | P value | No. of events | Events per 100 person-year | Hazard ratio (95% CI) | P value | |
| 1 | 1,147 | 8.8 | 1.00 | 1,147 | 8.9 | 1.00 | ||
| 2 | 171 | 16.5 | 1.37 (1.07–1.74) | 0.011 | 171 | 16.6 | 1.37 (1.07–1.74) | 0.011 |
| 3 | 43 | 25.8 | 1.35 (1.13–1.63) | < 0.001 | 43 | 26.0 | 1.36 (1.14–1.63) | < 0.001 |
| 4 | 25 | 46.1 | 1.24 (1.02–1.58) | 0.029 | 25 | 46.4 | 1.25 (1.03–1.51) | 0.026 |
Adjusted for baseline characteristics, such as age, sex, estimated glomerular filtration rate, albuminuria, hemoglobin, serum albumin, hemoglobin A1c, total cholesterol, and low-density lipoprotein cholesterol.
CI = confidence interval.
DISCUSSION
As shown in previous studies, SGLT2i use slowed kidney disease progression compared with nonuse in this cohort. Interestingly, heterogeneous phenotypes, which were linked to varying kidney outcomes, were identified among SGLT2i users. The rapidly deteriorating eGFR pattern was primarily observed in young patients, as well as those with significant albuminuria. This study is the first to identify these heterogeneous eGFR patterns following SGLT2i use and to explore the risk factors and outcomes associated with these patterns. These findings may help in the early identification of individuals at high risk of kidney disease progression after SGLT2i use.
The natural course of DKD exhibits significant heterogeneity and is influenced by various clinical and pathophysiological factors.11,17,18,23,24,25 In our cohort, despite SGLT2i use, a subset of patients experienced persistent declines in eGFR and increasing albuminuria, indicating nonresponsiveness to SGLT2is. Patients with severe albuminuria may already have advanced kidney damage, making recovery difficult and leading to faster progression of DKD.26,27 This outcome is likely because longstanding persistently high albuminuria levels significantly contribute to structural kidney deterioration, accelerating the decline in kidney function.25,28,29,30
Anemia is widely recognized as a risk factor for accelerated eGFR decline.31 The fact that anemia is also a complication of CKD and can present as a clinical sign of tubulointerstitial fibrosis suggests the possibility of a reduced response to SGLT2i.32,33 Additionally, anemia itself may negatively affect diabetic kidneys through hypoxia, even in the absence of structural damage.34 However, our study did not demonstrate a significant association between anemia and the declining trajectory.
Within the low baseline eGFR group, a higher baseline eGFR was associated with increased odds of belonging to the declining group. This association observed in the low baseline eGFR group may be explained by several hypotheses. First, patients with relatively higher eGFR values within a low baseline group may demonstrate greater absolute declines over time due to having a wider range for potential deterioration. Second, patients with very low baseline eGFR may receive more intensive clinical monitoring or earlier interventions, which could attenuate further decline. Lastly, bias may exist; patients with more advanced kidney dysfunction at baseline may have been censored earlier due to death or ESKD progression, resulting in a selectively stable subgroup remaining in the analysis.
Despite its strengths, such as its sufficient sample size for analysis and real-world context, this study has several limitations. This study was retrospective in design, which inherently restricts the ability to establish causal relationships. Unmeasured confounders likely influence the results, which could lead to a biased interpretation. The small number of patients in classes 3 and 4 identified in our study may limit their representativeness, highlighting the need for further validation in larger cohorts. In addition, the equation used to convert PCR to ACR has not been validated in the Korean population.22 Variations in clinical practices, patient demographics, and health care access across different centers could lead to different outcomes; thus, multicenter studies would be necessary to validate these findings broadly.
This observational study supports potential renal benefits of SGLT2is in attenuating the progression of kidney disease and reducing mortality risks in diabetic patients. The presence of distinct eGFR trajectory patterns suggests that not all patients respond uniformly, with some continuing to experience a rapid decline in kidney function. Additionally, we found clear differences in long-term clinical outcomes among the trajectory groups based on the degree of eGFR decline. These findings underscore the importance of early identification and appropriate warning of high-risk patients, as well as the need for additional treatment strategies. Further research is needed to better understand the mechanisms underlying these heterogeneous phenotypes and to develop more effective individualized therapeutic approaches.
Footnotes
Disclosure: The authors have no potential conflicts of interest to disclose.
- Conceptualization: Han SS.
- Data curation: Bae S.
- Formal analysis: Bae S, Yun D, Kim J.
- Investigation: Bae S, Yun D.
- Methodology: Han SS.
- Supervision: Park S, Kim YC, Kim DK, Oh KH, Joo KW, Kim YS, Han SS.
- Writing - original draft: Bae S.
- Writing - review & editing: Han SS.
SUPPLEMENTARY MATERIALS
Baseline characteristics of patients
Baseline characteristics of patients after propensity score matching
Study outcomes in the group treated with sodium-glucose cotransporter 2 inhibitors compared with the group that did not use these inhibitors
Metrics of trajectory models
Kaplan–Meier curves showing the rates of (A) kidney and (B) composite outcomes according to the use of SGLT2i.
Trajectory of eGFRs per each patient stratified by the class.
References
- 1.Sun H, Saeedi P, Karuranga S, Pinkepank M, Ogurtsova K, Duncan BB, et al. IDF diabetes atlas: global, regional and country-level diabetes prevalence estimates for 2021 and projections for 2045. Diabetes Res Clin Pract. 2022;183:109119. doi: 10.1016/j.diabres.2021.109119. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Tuttle KR, Bakris GL, Bilous RW, Chiang JL, de Boer IH, Goldstein-Fuchs J, et al. Diabetic kidney disease: a report from an ADA Consensus Conference. Diabetes Care. 2014;37(10):2864–2883. doi: 10.2337/dc14-1296. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Mottl AK, Alicic R, Argyropoulos C, Brosius FC, Mauer M, Molitch M, et al. KDOQI US commentary on the KDIGO 2020 clinical practice guideline for diabetes management in CKD. Am J Kidney Dis. 2022;79(4):457–479. doi: 10.1053/j.ajkd.2021.09.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Afkarian M, Sachs MC, Kestenbaum B, Hirsch IB, Tuttle KR, Himmelfarb J, et al. Kidney disease and increased mortality risk in type 2 diabetes. J Am Soc Nephrol. 2013;24(2):302–308. doi: 10.1681/ASN.2012070718. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Neuen BL, Young T, Heerspink HJL, Neal B, Perkovic V, Billot L, et al. SGLT2 inhibitors for the prevention of kidney failure in patients with type 2 diabetes: a systematic review and meta-analysis. Lancet Diabetes Endocrinol. 2019;7(11):845–854. doi: 10.1016/S2213-8587(19)30256-6. [DOI] [PubMed] [Google Scholar]
- 6.Herrington WG, Staplin N, Wanner C, Green JB, Hauske SJ, Emberson JR, et al. Empagliflozin in patients with chronic kidney disease. N Engl J Med. 2023;388(2):117–127. doi: 10.1056/NEJMoa2204233. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Heerspink HJL, Stefánsson BV, Correa-Rotter R, Chertow GM, Greene T, Hou FF, et al. Dapagliflozin in patients with chronic kidney disease. N Engl J Med. 2020;383(15):1436–1446. doi: 10.1056/NEJMoa2024816. [DOI] [PubMed] [Google Scholar]
- 8.Stevens PE, Ahmed SB, Carrero JJ, Foster B, Francis A, Hall RK, et al. KDIGO 2024 clinical practice guideline for the evaluation and management of chronic kidney disease. Kidney Int. 2024;105(4S):S117–S314. doi: 10.1016/j.kint.2023.10.018. [DOI] [PubMed] [Google Scholar]
- 9.Selby NM, Taal MW. An updated overview of diabetic nephropathy: diagnosis, prognosis, treatment goals and latest guidelines. Diabetes Obes Metab. 2020;22(S1):3–15. doi: 10.1111/dom.14007. [DOI] [PubMed] [Google Scholar]
- 10.Afkarian M, Zelnick LR, Hall YN, Heagerty PJ, Tuttle K, Weiss NS, et al. Clinical manifestations of kidney disease among US adults with diabetes, 1988-2014. JAMA. 2016;316(6):602–610. doi: 10.1001/jama.2016.10924. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Oshima M, Shimizu M, Yamanouchi M, Toyama T, Hara A, Furuichi K, et al. Trajectories of kidney function in diabetes: a clinicopathological update. Nat Rev Nephrol. 2021;17(11):740–750. doi: 10.1038/s41581-021-00462-y. [DOI] [PubMed] [Google Scholar]
- 12.Penno G, Solini A, Bonora E, Fondelli C, Orsi E, Zerbini G, et al. Clinical significance of nonalbuminuric renal impairment in type 2 diabetes. J Hypertens. 2011;29(9):1802–1809. doi: 10.1097/HJH.0b013e3283495cd6. [DOI] [PubMed] [Google Scholar]
- 13.D'Marco L, Guerra-Torres X, Viejo I, Lopez-Romero L, Yugueros A, Bermidez V. Non-albuminuric diabetic kidney disease phenotype: beyond albuminuria. touchREV Endocrinol. 2022;18(2):102–105. doi: 10.17925/EE.2022.18.2.102. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Warren B, Rebholz CM, Sang Y, Lee AK, Coresh J, Selvin E, et al. Diabetes and trajectories of estimated glomerular filtration rate: a prospective cohort analysis of the atherosclerosis risk in communities study. Diabetes Care. 2018;41(7)(8):1646–1653. doi: 10.2337/dc18-0277. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Krolewski AS, Skupien J, Rossing P, Warram JH. Fast renal decline to end-stage renal disease: an unrecognized feature of nephropathy in diabetes. Kidney Int. 2017;91(6):1300–1311. doi: 10.1016/j.kint.2016.10.046. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Chapter 2: definition, identification, and prediction of CKD progression. Kidney Int Suppl (2011) 2013;3(1):63–72. doi: 10.1038/kisup.2012.65. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Jiang G, Luk AOY, Tam CHT, Xie F, Carstensen B, Lau ESH, et al. Progression of diabetic kidney disease and trajectory of kidney function decline in Chinese patients with type 2 diabetes. Kidney Int. 2019;95(1):178–187. doi: 10.1016/j.kint.2018.08.026. [DOI] [PubMed] [Google Scholar]
- 18.Vistisen D, Andersen GS, Hulman A, Persson F, Rossing P, Jørgensen ME. Progressive decline in estimated glomerular filtration rate in patients with diabetes after moderate loss in kidney function-even without albuminuria. Diabetes Care. 2019;42(10):1886–1894. doi: 10.2337/dc19-0349. [DOI] [PubMed] [Google Scholar]
- 19.Feng L, Bee YM, Fu X, Kwek JL, Chan CM, Jafar TH. Kidney function trajectories, associated factors, and outcomes in multiethnic Asian patients with type 2 diabetes. J Diabetes. 2024;16(9):e13523. doi: 10.1111/1753-0407.13523. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Klen J, Dolžan V. Treatment response to SGLT2 inhibitors: from clinical characteristics to genetic variations. Int J Mol Sci. 2021;22(18):9800. doi: 10.3390/ijms22189800. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Levey AS, Stevens LA, Schmid CH, Zhang YL, Castro AF, 3rd, Feldman HI, et al. A new equation to estimate glomerular filtration rate. Ann Intern Med. 2009;150(9):604–612. doi: 10.7326/0003-4819-150-9-200905050-00006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Weaver RG, James MT, Ravani P, Weaver CGW, Lamb EJ, Tonelli M, et al. Estimating urine albumin-to-creatinine ratio from protein-to-creatinine ratio: development of equations using same-day measurements. J Am Soc Nephrol. 2020;31(3):591–601. doi: 10.1681/ASN.2019060605. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Davis TM, Chubb SA, Davis WA. The relationship between estimated glomerular filtration rate trajectory and all-cause mortality in type 2 diabetes: the fremantle diabetes study. Eur J Endocrinol. 2016;175(4):273–285. doi: 10.1530/EJE-16-0327. [DOI] [PubMed] [Google Scholar]
- 24.Weldegiorgis M, de Zeeuw D, Li L, Parving HH, Hou FF, Remuzzi G, et al. Longitudinal estimated GFR trajectories in patients with and without type 2 diabetes and nephropathy. Am J Kidney Dis. 2018;71(1):91–101. doi: 10.1053/j.ajkd.2017.08.010. [DOI] [PubMed] [Google Scholar]
- 25.Furuichi K, Yuzawa Y, Shimizu M, Hara A, Toyama T, Kitamura H, et al. Nationwide multicentre kidney biopsy study of Japanese patients with type 2 diabetes. Nephrol Dial Transplant. 2018;33(1):138–148. doi: 10.1093/ndt/gfw417. [DOI] [PubMed] [Google Scholar]
- 26.Yamanouchi M, Sawa N, Toyama T, Shimizu M, Oshima M, Yoshimura Y, et al. Trajectory of GFR decline and fluctuation in albuminuria leading to end-stage kidney disease in patients with biopsy-confirmed diabetic kidney disease. Kidney Int Rep. 2024;9(2):323–333. doi: 10.1016/j.ekir.2023.11.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Lachin JM, Nathan DM DCCT/EDIC Research Group. Understanding metabolic memory: the prolonged influence of glycemia during the diabetes control and complications trial (DCCT) on future risks of complications during the study of the epidemiology of diabetes interventions and complications (EDIC) Diabetes Care. 2021;44(10):2216–2224. doi: 10.2337/dc20-3097. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Shimizu M, Furuichi K, Toyama T, Kitajima S, Hara A, Kitagawa K, et al. Long-term outcomes of Japanese type 2 diabetic patients with biopsy-proven diabetic nephropathy. Diabetes Care. 2013;36(11):3655–3662. doi: 10.2337/dc13-0298. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Yasuda F, Mii A, Morita M, Aoki M, Tagawa M, Aratani S, et al. Importance of frequency and morphological characteristics of nodular diabetic glomerulosclerosis in diabetic nephropathy. Hum Pathol. 2018;75:95–103. doi: 10.1016/j.humpath.2018.01.019. [DOI] [PubMed] [Google Scholar]
- 30.Tonneijck L, Muskiet MH, Smits MM, van Bommel EJ, Heerspink HJ, van Raalte DH, et al. Glomerular hyperfiltration in diabetes: mechanisms, clinical significance, and treatment. J Am Soc Nephrol. 2017;28(4):1023–1039. doi: 10.1681/ASN.2016060666. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Xie L, Shao X, Yu Y, Gong W, Sun F, Wang M, et al. Anemia is a risk factor for rapid eGFR decline in type 2 diabetes. Front Endocrinol (Lausanne) 2023;14:1052227. doi: 10.3389/fendo.2023.1052227. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Nath KA. Tubulointerstitial changes as a major determinant in the progression of renal damage. Am J Kidney Dis. 1992;20(1):1–17. doi: 10.1016/s0272-6386(12)80312-x. [DOI] [PubMed] [Google Scholar]
- 33.Ito K, Yokota S, Watanabe M, Inoue Y, Takahashi K, Himuro N, et al. Anemia in diabetic patients reflects severe tubulointerstitial injury and aids in clinically predicting a diagnosis of diabetic nephropathy. Intern Med. 2021;60(9):1349–1357. doi: 10.2169/internalmedicine.5455-20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Rossert J, Froissart M. Role of anemia in progression of chronic kidney disease. Semin Nephrol. 2006;26(4):283–289. doi: 10.1016/j.semnephrol.2006.05.004. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Baseline characteristics of patients
Baseline characteristics of patients after propensity score matching
Study outcomes in the group treated with sodium-glucose cotransporter 2 inhibitors compared with the group that did not use these inhibitors
Metrics of trajectory models
Kaplan–Meier curves showing the rates of (A) kidney and (B) composite outcomes according to the use of SGLT2i.
Trajectory of eGFRs per each patient stratified by the class.