Initial effect of sodium-glucose cotransporter-2 inhibitors on kidney function in patients with immunoglobulin A nephropathy
Article information
Abstract
Background
Optimized supportive therapy, including maximal renin-angiotensin-aldosterone system (RAAS) inhibition, is not always adequate in patients with IgA nephropathy (IgAN). This study evaluated the early effects of sodium-glucose cotransporter-2 inhibitors (SGLT2is) on kidney function in patients with IgAN receiving RAAS inhibitors.
Methods
Adult patients who were newly diagnosed with IgAN during 2020–2024, with baseline estimated glomerular filtration rate (eGFR) ≥35 mL/min/1.73 m2, urine protein-to-creatinine ratio (UPCR) ≥0.8 g/g, and receiving RAAS inhibitors for ≥28 days were selected. The primary exposure was the incident use of SGLT2i. Changes in UPCR, urine albumin-to-creatinine ratio (UACR), and eGFR at 12 months were examined using mixed-effect models for repeated measures with inverse probability of treatment weighting.
Results
SGLT2i use was associated with an early eGFR decline (mean difference, –5.39 mL/min/1.73 m2; 95% confidence interval [CI], –9.51 to –1.26) but was not significantly associated with the reduction in UPCR (geometric mean ratio [GMR], 0.90; 95% CI, 0.68–1.18) or UACR (GMR, 0.94; 95% CI, 0.67–1.34) at 12 months compared with non-users. However, subgroup and sensitivity analyses suggested a potential relationship between SGLT2i use and the reduction in proteinuria in patients with S0 lesions according to the Oxford Classification and baseline systolic blood pressure ≥130 mmHg, indicating possible benefits despite the absence of consistent statistical significance across analyses.
Conclusion
In this real-world IgAN cohort, although SGLT2i use was not significantly associated with proteinuria reduction at 12 months, differential subgroup responses suggest that antiproteinuric benefits may be greater in patients without segmental glomerulosclerosis or with elevated systolic blood pressure.
Introduction
Immunoglobulin A nephropathy (IgAN) is the most common biopsy-confirmed primary glomerulonephritis worldwide. Recently, novel pharmacotherapeutic agents have been introduced to target the synthesis of pathogenic immunoglobulin A (IgA) and formation of IgA-containing immune complexes (IgA-IC), aiming to prevent IgA/IgA-IC–mediated kidney injury [1]. However, clinical trials do not provide equal opportunities to all patients, and systemic glucocorticoid therapy—the conventional treatment targeting IgAN-specific drivers of nephron loss—can cause significant adverse effects [2]. Consequently, optimized supportive therapy, including strict blood pressure control with maximization of renin-angiotensin-aldosterone system (RAAS) inhibition and lifestyle modification, remains the cornerstone of IgAN management [3]. Despite these interventions, treatment outcomes are often suboptimal, leaving a considerable proportion of patients at high risk of disease progression and highlighting the need for additional therapeutic strategies.
In the Dapagliflozin and Prevention of Adverse Outcomes in Chronic Kidney Disease (DAPA-CKD) trial, dapagliflozin treatment significantly reduced the incidence of primary outcome events, leading to early trial termination due to efficacy [4]. In a prespecified subgroup of 270 participants with IgAN in the DAPA-CKD trial, dapagliflozin was associated with a significant reduction in major adverse kidney outcomes and decreased albuminuria compared with placebo [5]. This aligns with large-scale randomized controlled trials demonstrating the renoprotective effects of sodium-glucose cotransporter-2 inhibitors (SGLT2is) in diabetic and non-diabetic CKD populations, suggesting that SGLT2i can be considered as an adjunct to standard care in IgAN. However, given that the risk of kidney function decline in IgAN is highly heterogeneous, with the 10-year risk of end-stage kidney disease ranging from 5% to 60% [6], the kidney-protective effects of SGLT2i may not be uniform across all patients with IgAN. Moreover, real-world evidence remains limited regarding which IgAN subpopulation, particularly those already receiving standard RAAS inhibition, would derive the greatest benefit from SGLT2i therapy.
Therefore, this study aimed to assess the association between SGLT2i use and early changes in kidney function among patients with newly diagnosed IgAN receiving RAAS inhibitors. In a multicenter cohort of biopsy-proven patients, kidney outcomes—primarily proteinuria reduction—over a 12-month period were assessed. Subgroup analyses, including stratification by histological scores according to the Oxford Classification, were performed to identify patient characteristics associated with the greatest benefit from SGLT2i therapy.
Methods
Study population
Adult patients (aged ≥ 18 years) with newly diagnosed, biopsy-confirmed IgAN at Severance Hospital and National Health Insurance Service Ilsan Hospital between January 1, 2020, and December 31, 2024, were initially screened. Patients with secondary forms of IgAN, non-IgAN glomerulonephritis, or a history of polycystic kidney disease were considered ineligible for inclusion in the study. The date of kidney biopsy was defined as the baseline. Eligible participants met the following criteria: baseline estimated glomerular filtration rate (eGFR) ≥35 mL/min/1.73 m2, calculated using the CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration) equation [7]; baseline proteinuria, defined as a spot urine protein-to-creatinine ratio (UPCR) ≥0.8 g/g; and treatment with RAAS inhibitors—an angiotensin-converting enzyme inhibitor and/or an angiotensin II type I receptor blocker—for at least 28 days within the designated window period (n = 224). Information on RAAS inhibitor dosing and tolerability-related factors relevant to dose maintenance or further up-titration, including hyperkalemia, orthostatic hypotension, and a clinically relevant decline in eGFR, was obtained through review of the medical records. The window period was defined as a 3-month timeframe from the date of kidney biopsy. Patients with diabetes were included in the study cohort if their diabetes was adequately controlled, defined as a glycated hemoglobin level of ≤8% (64 mmol/mol). The key exclusion criteria were prior dialysis treatment or kidney transplantation (n = 44), systemic glucocorticoid or immunosuppressant use within 180 days before baseline (n = 6), and fewer than two confirmed measurements of proteinuria and eGFR during follow-up (n = 40). To ensure that only new users were included, it was confirmed that none of the participants included in the study had previously used SGLT2is in the year preceding baseline. The final cohort comprised 134 participants (Supplementary Fig. 1, available online). Patients were followed up until the initiation of immunosuppressive treatment, dialysis, kidney transplantation, death, or 1 year from baseline.
The study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Review Boards of Severance Hospital (IRB No. 4-2024-1381) and National Health Insurance Service Ilsan Hospital (IRB No. NHIMC-2024-04-003). Given the nonintrusive nature of this study and patient anonymity, the requirement for written consent was waived.
Data collection
Baseline demographic and clinical data were collected during a 180-day look-back period, with follow-up data obtained at 90-day intervals. Comorbid diabetes mellitus and heart failure were identified using ICD-10 (International Statistical Classification of Diseases and Related Health Problems, 10th Revision) codes, and cross-checked with glucose-lowering medications for diabetes. Medication use, including RAAS inhibitors, diuretics, and statins, was reviewed. Urine tests were performed on fresh samples using a regularly calibrated semiautomatic urine analyzer. When two or more laboratory measurements were available within each 90-day study interval, the earliest measured value was used. Blood pressure was measured during clinic visits using the auscultatory or oscillometric method. Detailed descriptions of the data collection are provided in Supplementary Method 1 (available online).
Kidney biopsy findings were scored according to the Oxford Classification: mesangial hypercellularity, M0/M1 (≤50%/>50% of glomeruli with ≥4 mesangial cells per area); endocapillary hypercellularity, E0/E1 (absent/present); segmental glomerulosclerosis, S0/S1 (absent/present); tubular atrophy/interstitial fibrosis, T0/T1 or T2 (≤25%/26%–50% or >50%); cellular or fibrocellular crescents, C0/C1 or C2 (no crescents/crescents in at least 1 but <25% of glomeruli or crescents in at least 25% of glomeruli) [8,9].
Exposure and outcome ascertainment
The exposure of interest was incident SGLT2i use. Patients prescribed SGLT2i for at least 28 days within 90 days of the kidney biopsy (window period) were classified as the treatment group. The primary outcome was the change in UPCR at 12 months, whereas secondary outcomes included changes in urine albumin-to-creatinine ratio (UACR) and eGFR at 12 months.
Statistical analysis
Continuous variables are reported as mean ± standard deviation or median (interquartile range [IQR]), and categorical variables are expressed as number (percentage). UPCR and UACR were log-transformed before analysis. Inverse probability of treatment weighting (IPTW) was used to balance the baseline characteristics between SGLT2i users and non-users. Stabilized weights were estimated for each participant using logistic regression-based propensity scores derived from baseline characteristics of age, sex, systolic blood pressure (SBP), diastolic blood pressure (DBP), history of diabetes, use of diuretics and statins, eGFR, UPCR, and scores from each component of the Oxford Classification from the pathological diagnosis. To prevent potential violation of the positivity assumption, stabilized weights were truncated below the 1st or above the 99th percentile. Standardized mean differences (SMDs) were calculated to compare the balance of baseline covariates between the two groups.
Changes in UPCR and UACR were evaluated using a weighted mixed-effects model for repeated measures (MMRM), which included data from 3, 6, 9, and 12 months. The model incorporated terms for fixed effects for treatment group, time (corresponding to 3, 6, 9, and 12 months), and their interactions and was adjusted for the same covariates used in the IPTW model. From the fitted MMRM, least-squares mean UPCR and UACR values (geometric means) were obtained for each treatment group by back-transforming the adjusted means from the log scale. Between-group differences in UPCR and UACR at 12 months are expressed as geometric mean ratios (GMRs) with corresponding 95% confidence intervals (CIs), derived from the ratio of the geometric least-squares means.
Time-specific mean differences in eGFR were estimated using IPTW-weighted MMRM in parallel with the primary analysis and are reported as between-group least-squares mean eGFR differences with 95% CIs at 12 months. The overall 1-year eGFR slope was estimated using a weighted random coefficient model.
Subgroup analyses for UPCR and UACR were performed based on the Oxford Classification components and clinical characteristics, including age (<50 years vs. ≥50 years), sex, diabetes mellitus (presence vs. absence), SBP (<130 mmHg vs. ≥130 mmHg), eGFR (<45 mL/min/1.73 m2 vs. ≥45 mL/min/1.73 m2), and UPCR (<1.5 g/g vs. ≥1.5 g/g). An additional sensitivity analysis was conducted using overlap weighting, which emphasizes patients in clinical equipoise (propensity scores near 0.5), thereby targeting the average treatment effect in the overlap population [10]. Detailed descriptions of the MMRM and sensitivity analyses are provided in Supplementary Method 2 (available online). Statistical significance was set at p < 0.05. All analyses were performed using SAS (version 9.4, SAS Institute) and R (version 4.3.3, R Foundation for Statistical Computing).
Results
Baseline characteristics of study participants
The baseline characteristics of the patients are summarized in Table 1. The mean age was 45.8 ± 13.6 years, and 34.3% of the patients were men. Among the study participants, 10.4% had a history of diabetes, and only one patient (non-user of SGLT2i) experienced heart failure. Patients treated with SGLT2i tended to be younger, more frequently male, and had higher DBP; they were more likely to be prescribed statins. Among SGLT2i users, the median baseline UPCR was 1,280.0 mg/g (IQR, 1,060.0–2,570.0 mg/g), and the mean eGFR was 75.4 ± 24.8 mL/min/1.73 m2, whereas corresponding values among non-users were 1,315.0 mg/g (IQR, 960.4–1,790.0 mg/g) and 73.9 ± 25.1 mL/min/1.73 m2. Regarding the Oxford Classification, SGLT2i users had higher M, E, S, and T scores and lower C scores than non-users. After IPTW adjustment, the SMDs of the baseline characteristics were well balanced between the two groups.
Baseline characteristics between study groups
Association between sodium-glucose cotransporter-2 inhibitor use and study outcomes
Of participants included in the primary analysis, 91 (83.5%) completed follow-up through the 12-month study period. After 12 months, patients who received SGLT2i were associated with a 10% reduction in UPCR compared with non-users; however, this between-group difference was not statistically significant (GMR, 0.90; 95% CI, 0.68–1.18; p = 0.44). The findings for UACR demonstrated a similar pattern, although the between-group difference did not reach statistical significance (GMR, 0.94; 95% CI, 0.67–1.34; p = 0.74). In contrast, SGLT2i use was associated with a greater decline in eGFR, with a between-group mean difference of –5.39 mL/min/1.73 m2 (95% CI, –9.51 to –1.26 mL/min/1.73 m2; p = 0.01) compared with that in non-users. The 1-year eGFR slope difference was translated as –5.58 mL/min/1.73 m2 per year (p = 0.03) between the two groups.
Subgroup analyses
In the prespecified subgroup analyses stratified by each component of the Oxford Classification, SGLT2i use was associated with a reduction in UPCR at 12 months among patients classified with S0 lesions (GMR, 0.42; 95% CI, 0.18–0.95), although the inter-subgroup difference was not evident (GMR, 0.97; 95% CI, 0.73–1.29 in the subgroup of patients classified with S1) (Fig. 1). No significant association between SGLT2i therapy and UPCR changes was demonstrated in the other Oxford Classification-defined subgroups.
Geometric LS means of UPCR at 12 months compared with those at baseline across subgroups stratified by Oxford Classification.
CI, confidence interval; LS, least squares; SGLT2i, sodium-glucose cotransporter-2 inhibitor; UPCR, urine protein-to-creatinine ratio.
In analyses stratified by additional clinical factors, a reduction in UPCR at 12 months associated with SGLT2i use was observed in patients with baseline SBP ≥130 mmHg (GMR, 0.45; 95% CI, 0.20–0.98) with the absence of effect modification, as the GMR for those with SBP <130 mmHg was 1.13 (95% CI, 0.85–1.50) (Fig. 2). Likewise, the remaining subgroup analyses did not demonstrate statistically significant associations between SGLT2i use and UPCR changes at 12 months in the other clinical subgroups.
Geometric LS means of UPCR at 12 months compared with those at baseline across predefined subgroups.
CI, confidence interval; eGFR, estimated glomerular filtration rate; LS, least squares; SBP, systolic blood pressure; SGLT2i, sodium-glucose cotransporter-2 inhibitor; UPCR, urine protein-to-creatinine ratio.
Analyses using UACR outcomes showed results broadly consistent with those using UPCR outcomes. SGLT2i use was associated with a reduction in UACR at 12 months among patients categorized as having S0 lesions according to the Oxford Classification (GMR, 0.38; 95% CI, 0.16–0.92); however, this association did not reach statistical significance when compared with the subgroup diagnosed as S1 (GMR, 1.02; 95% CI, 0.71–1.48). In contrast, effect modification was identified in the subgroup analysis based on the SBP. Compared with patients having SBP <130 mmHg (GMR, 1.28; 95% CI, 0.91–1.79), SGLT2i use was significantly associated with UACR reduction at 12 months among patients with SBP ≥130 mmHg (GMR, 0.21; 95% CI, 0.05–0.84) (Fig. 3).
Geometric LS means of UACR at 12 months compared with those at baseline across subgroups.
CI, confidence interval; eGFR, estimated glomerular filtration rate; LS, least squares; SBP, systolic blood pressure; SGLT2i, sodium-glucose cotransporter-2 inhibitor; UPCR, urine protein-to-creatinine ratio.
Sensitivity analysis
In the additional sensitivity analysis using overlap weighting, the results were consistent with those of the primary analyses. The GMR for UPCR reduction among SGLT2i users was 0.95 (95% CI, 0.74–1.23) compared with that in non-users. When the assessments were repeated within the prespecified subgroups, similar results were observed: SGLT2i use was significantly associated with UPCR reduction at 12 months in patients with S0 lesions in the Oxford Classification (GMR, 0.36; 95% CI, 0.17–0.76), when compared with patients with S1 (GMR, 1.03; 95% CI, 0.78–1.36). In the analyses stratified by SBP, the GMR among patients with SBP ≥130 mmHg was 0.50 (95% CI, 0.28–0.89) (Fig. 4).
Sensitivity analysis: geometric LS means of UPCR at 12 months compared with those at baseline across subgroups using overlap weighting.
CI, confidence interval; eGFR, estimated glomerular filtration rate; LS, least squares; SBP, systolic blood pressure; SGLT2i, sodium-glucose cotransporter-2 inhibitor; UPCR, urine protein-to-creatinine ratio.
Discussion
In this real-world cohort study of patients with biopsy-confirmed IgAN receiving standard care, no significant association was observed between SGLT2i use and proteinuria reduction at 12 months; however, notable signals emerged in subgroup analyses, particularly among patients with S0 lesions according to the Oxford Classification and those with elevated baseline SBP. Given the retrospective nature of this study, the limited sample size, and the potentially insufficient follow-up duration to evaluate the long-term effects of SGLT2i, to the best of our knowledge, no previous study has investigated whether the renoprotective efficacy of SGLT2i varies according to specific clinical or histological characteristics among patients with IgAN receiving standard care.
The current study identified an early decline in eGFR following the initiation of SGLT2i therapy in patients with IgAN. This decrease is similar to the reversible “eGFR dip” previously described in earlier studies [11], which is considered to be a transient hemodynamic response rather than true deterioration in kidney function. The anticipated subsequent stabilization of eGFR following this initial dip could not be confirmed within the 12-month follow-up period. However, this study did not observe a statistically significant correlation between SGLT2i use and the reduction of proteinuria at 12 months in the overall cohort. Given that persistent proteinuria serves as a predictive factor for renal function decline and increased mortality in IgAN [6], reducing proteinuria remains a critical therapeutic objective. The disparity between the findings of the current study and those of the prespecified DAPA-CKD subgroup analysis, where dapagliflozin reduced UACR by 26% over a median 2.1-year follow-up in patients with IgAN [5], may be attributable to several factors. First, the 12-month follow-up period of the current study may not have been long enough to fully capture the long-term antiproteinuric and renoprotective effects of SGLT2i. Moreover, the initial hemodynamic shift accompanying the early eGFR dip may transiently influence filtration dynamics, potentially masking a gradual antiproteinuric trend within this limited observation window. In a previous short-term randomized crossover trial of non-diabetic patients with CKD, approximately half of whom had IgAN, dapagliflozin significantly reduced the measured GFR without producing a meaningful reduction in proteinuria compared with placebo [12], a pattern consistent with the observations of the current study. Second, the antiproteinuric effect of SGLT2i in IgAN may vary depending on disease activity. In comparison with the current study, which included patients with recent kidney biopsy-confirmed incident IgAN, the DAPA-CKD trial did not report the timing of IgAN diagnosis or disease duration. In cases of diabetic kidney disease, SGLT2i has been proposed to reduce proteinuria through restoration of tubuloglomerular feedback, reduction of glomerular hyperfiltration, and attenuation of intraglomerular pressure [13,14]. However, the effect of SGLT2i on the regulation of proteinuria in IgAN may be influenced by inflammatory or immunomodulatory processes that vary across the disease course. Third, given the relatively small sample size of this study, statistical power may have been limited to detect a modest antiproteinuric effect of SGLT2i. Therefore, the possibility of a type II error cannot be excluded, and statistically nonsignificant findings should be interpreted with caution.
The subgroup and sensitivity analyses in this study offer important insights into the clinical contexts in which SGLT2i therapy may exert the greatest antiproteinuric benefit in IgAN. A notable observation was the differential effect according to the presence of segmental glomerulosclerosis, a marker of chronic and irreversible injury, although statistical evidence for effect modification across subgroups was not consistently demonstrated. Previous studies have shown that segmental glomerulosclerosis in IgAN is associated with greater proteinuria at the time of biopsy, a higher risk of disease progression, and diminished therapeutic responsiveness [15–19]. Glomeruli that are relatively intact (S0) can be more amenable to hemodynamic and antiproteinuric interventions than those with fixed structural defects (S1), thus allowing agents such as SGLT2i to reduce intraglomerular pressure more effectively. In other words, the benefits of SGLT2 inhibition may be pronounced in patients earlier in the disease course, with more functional and structural reserve, represented by S0 status. Nonetheless, given the small number of SGLT2i users with S0 lesions and the limited robustness of these subgroup findings, the observed association should be interpreted with caution and considered to be hypothesis-generating only, rather than as evidence of a definite clinical benefit of SGLT2i in this specific subpopulation. In addition, future mechanistic studies with larger cohorts are warranted to investigate the role of the Oxford Classification as a pathological predictor of SGLT2i responsiveness in IgAN.
Another positive signal was that SGLT2i efficacy appeared to depend on SBP, with patients having baseline SBP ≥130 mmHg showing an association with proteinuria reduction. Although the analysis of the primary outcome of UPCR did not reach statistical significance when compared with those with SBP <130 mmHg, the effect modification became more apparent when evaluating UACR as a secondary outcome. In the DAPA-CKD trial, dapagliflozin lowered blood pressure by approximately 3–4 mmHg during follow-up [20], an effect thought to stem mainly from its natriuretic and diuretic properties. In patients with hypertensive IgAN, heightened hemodynamic stress can lead to persistently elevated intraglomerular pressure that may not be fully mitigated by RAAS inhibition alone, partly due to angiotensin and/or aldosterone escape [21,22]. Consequently, adding an SGLT2i to RAAS blockade may provide further glomerular “unloading,” potentially leading to additional reduction in proteinuria in this subgroup. However, caution must be exercised when interpreting this subgroup finding, because the possible contribution of incomplete optimization of background supportive care due to tolerability-related limitations, such as hyperkalemia or orthostatic hypotension, cannot be fully excluded.
The findings of this study should be interpreted with caution due to certain limitations. First, causality cannot be inferred due to the observational study design, and the possibility of residual bias due to potential unmeasured confounders cannot be entirely ruled out. In particular, SGLT2i use was evaluated in an “early treatment” context shortly after IgAN diagnosis, concurrently with conventional supportive therapy, which alone could substantially alleviate the disease burden. The degree to which the entire cohort received optimized supportive care, including lifestyle or dietary interventions, could not be fully accounted for due to limited information. Although a formal run-in period of stable RAAS inhibition prior to SGLT2i exposure could not be implemented in this cohort of biopsy-confirmed newly diagnosed IgAN, all included patients received RAAS inhibitors, a key pillar of supportive care for IgAN, for at least 28 days within the designated window period. Moreover, analyses incorporated propensity scores and adjusted for clinical factors and histological characteristics to account for variations in treatment decisions and responsiveness. In addition, patients who received immunotherapy, including glucocorticoids, before enrollment were excluded, and those who initiated immunosuppression as a rescue medication during follow-up were censored to minimize selection bias related to disease severity and confounding by indication. Second, the sample size and number of patients on SGLT2is were modest compared with those of large-scale randomized trials. However, the number of participants in each group was sufficient for the modelling approach, and sensitivity analysis using overlap weighting, which is better suited to estimate the treatment effects in subgroups with clinical equipoise, yielded consistent results, suggesting that the relatively small cohort size is unlikely to fully negate the observed associations. Third, proteinuria was assessed using spot UPCR rather than 24-hour collections, which may introduce measurement variability, although UPCR is widely accepted in clinical practice. Finally, the study population consisted exclusively of Korean participants, which may limit the generalizability of these findings to patients with IgAN in other populations.
In conclusion, although the early eGFR decline observed in SGLT2i users aligns with previous findings, compelling evidence regarding the impact of SGLT2i on proteinuria reduction at 12 months remains inconclusive across the entire cohort of patients with IgAN derived from real-world data. However, differential responses in subgroup analyses suggest that the antiproteinuric benefit of SGLT2i may be enhanced in patients with specific characteristics, such as the absence of segmental glomerulosclerosis or elevated SBP. Given the limited evidence supporting a universal renoprotective effect of SGLT2i in all patients with IgAN, these findings highlight the potential for individualized therapy, with SGLT2i serving as a novel therapeutic adjunct to RAAS inhibition.
Supplementary Materials
Supplementary data are available at Kidney Research and Clinical Practice online (https://doi.org/10.23876/j.krcp.26.072).
Notes
Conflicts of interest
Tae-Hyun Yoo is the Editor-in-Chief of Kidney Research and Clinical Practice and was not involved in the review process of this article. All authors have no other conflicts of interest to declare.
Funding
This Work was supported by the National Health Insurance Service Ilsan Hospital grant (NHIMC-2024-CR-041). The funding source played no role in the study design, data collection, data analysis, decision to publish, or manuscript preparation.
Data sharing statement
All relevant data are within the manuscript and Supplemental Materials. Partial restrictions to the data and/or materials apply. The technical appendix and statistical codes are available from Dr. Jae Young Kim upon request (E-mail: kim.jy@nhimc.or.kr).
Authors’ contributions
Conceptualization: JYK, HWK
Data curation: JYK, JHK
Formal analysis, Funding acquisition, Project administration: JYK
Investigation: JYK, SEC
Supervision: All authors
Writing–original draft: JYK, EWK, HWK
Writing–review & editing: All authors
All authors read and approved the final manuscript.