Twenty-Four-Week UACR Response Trajectories and Clinical Correlates of Residual Albuminuria in Diabetic Kidney Disease During Finerenone Therapy: A Real-World Longitudinal Study Authors: Yang Zhang, Qin Zhu, Chaochao Wang, Xiaoqiao Cai, Shengfen Lin, Beibei Ding, Xingyu Zhu, Yongqiang Lin Published: 2026 ABSTRACT AbstractBackgroundAfter finerenone therapy, some patients with diabetic kidney disease (DKD) continue to have residual albuminuria. A single time-point change in urinary albumin-to-creatinine ratio (UACR) may not fully capture response heterogeneity in real-world practice. This study aimed to identify 24-week baseline-corrected UACR response trajectories in patients with DKD and residual albuminuria during finerenone therapy and to examine associated clinical factors.MethodsThis single-center retrospective real-world longitudinal cohort study included 136 patients with DKD who had residual albuminuria after at least 4 weeks of finerenone therapy and available longitudinal UACR data. Based on multi-time-point log(UACR_t/UACR_baseline) data over 24 weeks, latent class mixed models were used to identify UACR response trajectories and analyze clinical factors associated with the persistent residual albuminuria trajectory.ResultsThree UACR response trajectories were identified: favorable response in 43 patients (31.6%), partial response in 62 patients (45.6%), and persistent residual albuminuria in 31 patients (22.8%); the latter 2 patterns accounted for 68.4% of the selected residual-albuminuria cohort. The median relative changes in UACR at week 24 were −51.6%, −24.8%, and −4.9%, respectively. Higher HbA1c was associated with higher estimated odds of the persistent residual albuminuria trajectory (OR, 1.46; 95% CI, 1.07–2.00). The estimated directions for systolic blood pressure and sodium-glucose cotransporter 2 (SGLT2) inhibitor use were consistent, although uncertainty remained in extended or selected sensitivity analyses. Week-24 eGFR changes did not differ clearly across trajectories, and continuous UACR change was not clearly associated with eGFR change.ConclusionPatients with DKD and residual albuminuria during finerenone therapy showed distinct 24-week UACR response trajectories. Trajectory analysis provides exploratory real-world evidence for dynamic risk stratification, but the identified patterns and clinical correlates require validation in larger, prospective, multicenter cohorts with longer follow-up.Keywords: diabetic kidney disease, finerenone, residual albuminuria, trajectory analysis, real-world study Introduction Diabetic kidney disease (DKD) is a major cause of chronic kidney disease progression, end-stage kidney disease, and increased cardiovascular risk in patients with type 2 diabetes. Urinary albumin-to-creatinine ratio (UACR) and estimated glomerular filtration rate (eGFR) are core indicators for risk stratification, treatment monitoring, and prognostic assessment in DKD.1,2 Compared with eGFR, changes in UACR often provide an earlier signal of glomerular barrier injury, hemodynamic change, and treatment response. Previous individual participant data meta-analyses and surrogate endpoint assessments have shown that early reductions in albuminuria are closely associated with lower subsequent risk of composite kidney outcomes. These findings support UACR as an important intermediate outcome in short- to medium-term DKD risk management and clinical research.3 However, UACR is influenced by blood pressure, glycemic status, infection, acute changes in kidney function, sampling time, and adjustment of background therapy. A single time-point change may therefore be insufficient to describe dynamic response patterns during treatment.4,5 In recent years, DKD treatment has shifted from reliance on renin-angiotensin system inhibition alone toward a multi-mechanistic kidney-protective strategy. The use of SGLT2 inhibitors, glucagon-like peptide-1 (GLP-1) receptor agonists, and nonsteroidal mineralocorticoid receptor antagonists has substantially changed the management of chronic kidney disease associated with type 2 diabetes.6–9 Finerenone has been shown in FIDELIO-DKD, FIGARO-DKD, and the pooled FIDELITY analysis to reduce kidney and cardiovascular event risk and to produce early reductions in albuminuria.10–12 Nevertheless, patients in routine practice differ substantially in baseline kidney risk, concomitant treatment, blood pressure and glycemic control, treatment adherence, and inflammatory or fibrotic status. Even during finerenone therapy, a considerable proportion of patients continue to have residual albuminuria or an insufficient decline in albuminuria.13,14 Although fixed time-point UACR change or a binary definition of a UACR reduction of at least 30% is simple, such approaches may obscure distinct response patterns, including early rapid decline, sustained partial decline, fluctuation, and persistent residual albuminuria.15 Thus, trajectory analysis based on multi-time-point UACR changes relative to the individual baseline may better characterize heterogeneity in residual risk after finerenone therapy. Real-world evidence on baseline-corrected UACR response trajectories and their clinical correlates after finerenone therapy in patients with DKD remains limited.16 Therefore, this study enrolled patients with DKD and residual albuminuria during finerenone therapy. Using multi-time-point baseline-corrected UACR data over 24 weeks, we identified albuminuria response trajectories and analyzed their clinical correlates. We also described changes in urinary protein-to-creatinine ratio (UPCR), eGFR, serum potassium, and safety events. Use of Compound Centella Formula No. 2 and traditional Chinese medicine (TCM) syndrome-related information was described only as supplementary exploratory variables and was not used to infer the efficacy of Chinese herbal medicine or TCM syndrome mechanisms. This study aimed to provide preliminary real-world evidence for dynamic risk stratification in patients with residual albuminuria after finerenone therapy. Materials and Methods Study Design and Data Source This was a single-center retrospective real-world longitudinal cohort study designed to identify 24-week dynamic response trajectories of residual albuminuria in patients with DKD during finerenone therapy and to explore associated clinical factors. Patients were identified from the electronic medical record system of the nephrology outpatient and inpatient departments of our hospital. The index baseline dates ranged from January 1, 2024, to January 1, 2025, and follow-up data were collected through July 15, 2025. Study data included demographic information, diabetes and kidney disease history, comorbidities, medication records, laboratory tests, follow-up records, and adverse event records. The study was reported in accordance with the STROBE statement for observational studies. The RECORD statement was also considered in reporting routine clinical data sources, patient screening, missing data, and data-quality control. Reporting of trajectory models followed relevant GRoLTS principles.17–19 The study was approved by the Ethics Committee of our hospital (approval No. 2023-K049). Because this study used previously collected routine clinical data and did not involve additional interventions, the ethics committee waived the requirement for written informed consent. All data were de-identified before analysis. Study Population, Index Date, and Finerenone Treatment Timeline Patients were eligible if they were 18–80 years of age; had type 2 diabetes-related chronic kidney disease or DKD; had received finerenone for at least 4 weeks; had residual albuminuria during finerenone and individualized kidney-protective therapy; had an index baseline UACR of at least 30 mg/g; had an eGFR of at least 25 mL/min/1.73 m2; and had at least 3 valid UACR measurements within 24 weeks, including the index baseline measurement. Residual albuminuria was defined as a UACR of at least 30 mg/g during background kidney-protective therapy and finerenone treatment. DKD was diagnosed on the basis of diabetes history, persistent albuminuria, eGFR changes, and the overall clinical profile. Fundus findings and kidney biopsy findings were considered when necessary. In patients with suspected non-diabetic kidney disease, the disease course, laboratory findings, and pathological data were reviewed.20 Patients were excluded if they had acute kidney injury; end-stage kidney disease, maintenance dialysis, or previous kidney transplantation; another clearly defined cause of kidney disease; active infection, active malignant tumor, or severe liver dysfunction; pregnancy or lactation; use of glucocorticoids, immunosuppressants, or other treatments that could substantially affect the interpretation of proteinuria during the study period; UACR values clearly affected by urinary tract infection, fever, acute heart failure, acute kidney-function fluctuation, or recent strenuous exercise without an available alternative measurement; or severe missingness of key clinical data. The index baseline was defined as the clinical assessment date at which the patient still met the definition of residual albuminuria after finerenone initiation and had an available UACR record. This date was not necessarily the date of the first finerenone prescription. The interval from finerenone initiation to the index baseline, the index baseline dose, and dose adjustments during follow-up were extracted as treatment-timeline variables and are reported in the Results or Supplementary Materials. Clearly affected UACR values were excluded from the main analysis; a stricter sensitivity analysis additionally excluded patients with any follow-up window potentially influenced by an acute event. Background Therapy, UACR Follow-Up Windows, and Data-Quality Control The finerenone dose was determined by the treating physician according to eGFR, serum potassium level, and tolerability, and was usually 10 mg or 20 mg once daily.20 The study also recorded use of ACEI/ARB, SGLT2 inhibitors, GLP-1 receptor agonists, antihypertensive agents, glucose-lowering medications, lipid-lowering therapy, and other kidney-related treatments. Finerenone dose escalation, dose reduction, temporary interruption, and serum potassium monitoring were recorded. When first-prescription dates were sufficiently documented, ACEI/ARB and SGLT2 inhibitor use was categorized as before finerenone, after finerenone but before the index baseline, after the index baseline, never used, or unascertainable. Within each category, the number of patients with available week-24 UACR and the median relative change in UACR were summarized descriptively. No hypothesis testing or adjusted modeling was performed because several subgroups were small and treatment timing was nonrandom. UACR follow-up time points were extracted from real-world clinical records and mapped to the index baseline and the 4-, 8-, 12-, and 24-week windows. Each window allowed measurements within 2 weeks before or after the target time point. If multiple measurements were available within the same window, morning urine UACR was preferred. If morning urine was not available, the value closest to the target time point and not affected by acute clinical factors was used. UACR values clearly affected by urinary tract infection, fever, acute heart failure, acute kidney-function fluctuation, recent strenuous exercise, or other acute factors were not used in the main trajectory analysis.4,5 The number of available samples in each follow-up window, actual testing times, and missingness are reported in the Results and Supplementary Materials. Use of Compound Centella Formula No. 2 and TCM syndrome information were extracted only as exploratory real-world variables. Outcome Definitions, Safety Events, and Analysis Denominators The primary outcome was the 24-week baseline-corrected UACR response trajectory. The main analysis used the baseline-corrected log-UACR ratio, defined as log(UACR_t/UACR_baseline), to reduce the possibility that trajectory classification mainly reflected baseline UACR burden. Here, UACR_t denoted the UACR value within a follow-up window, and UACR_baseline denoted the index baseline UACR value. The log ratio at the index baseline was defined as 0, and subsequent follow-up values reflected dynamic UACR changes relative to each patient’s own baseline. Trajectory classes were used only to describe short-term albuminuria response patterns and were not used to infer treatment causality. Secondary outcomes included a UACR reduction of at least 30% from the index baseline at week 24, a UACR reduction of at least 50%, changes in UPCR, changes in eGFR, changes in HbA1c, and changes in systolic blood pressure. Other routine laboratory indicators and serum potassium tests were used mainly for background description, treatment monitoring, and safety assessment. Safety outcomes included hyperkalemia, severe hyperkalemia, elevated alanine aminotransferase (ALT) or aspartate aminotransferase (AST), leukopenia, thrombocytopenia, hemoglobin decrease, gastrointestinal discomfort, all-cause hospitalization, dialysis initiation, death, and definite treatment-related serious adverse events. Hyperkalemia was defined as serum potassium of at least 5.5 mmol/L, and severe hyperkalemia was defined as serum potassium of at least 6.0 mmol/L. ALT or AST elevation was defined as a value greater than 3 times the upper limit of normal. Leukopenia was defined as a white blood cell count less than 3.0 × 109/L, thrombocytopenia as a platelet count less than 100 × 109/L, and hemoglobin decrease as a decline greater than 20 g/L from the index baseline. Laboratory safety events were analyzed using patients with the relevant follow-up test as the denominator. Gastrointestinal discomfort, all-cause hospitalization, dialysis initiation, death, and definite treatment-related serious adverse events were analyzed using all included patients as the denominator. All safety results are described as n/N (%). Baseline-Corrected UACR Response Trajectory Modeling Latent class mixed models were used to identify 24-week baseline-corrected UACR response trajectories.19,21 Time was handled according to the actual follow-up week, and linear and quadratic time terms were compared during model development. The main model used an individual-level random intercept to account for within-patient correlation. Alternative specifications, including a random slope, were evaluated in supplementary diagnostics. Model fitting began with a 1-class model, followed by stepwise comparison of 2-, 3-, and 4-class models. Final model selection considered the Bayesian information criterion, Akaike information criterion, entropy, average posterior classification probability, sample proportion in each trajectory class, model convergence, and clinical interpretability. A trajectory class was not retained in the final model if it had an excessively small sample proportion, insufficient average posterior classification probability, or no stable clinical interpretation. Each patient was assigned to the most likely trajectory class according to the maximum posterior probability. Trajectories were named according to their week-24 relative UACR change patterns and clinical interpretability. The time-term form, random-effects specification, alternative model comparisons, starting-value procedures, convergence status, posterior classification probability distribution, and model diagnostics are reported in the Supplementary Materials. Raw log-UACR level trajectories were used as a sensitivity analysis to distinguish albuminuria burden trajectories from baseline-corrected response trajectories. Trajectory classes were considered statistical summaries and were not assumed to represent stable biological subtypes.19 Statistical Analysis and Sensitivity Analyses Continuous variables are presented as mean ± standard deviation or median (interquartile range), depending on their distribution. Categorical variables are presented as counts and percentages. Baseline characteristics across trajectory groups were compared mainly using descriptive statistics and effect-size interpretation, with P values used only as supportive information. Supplementary tables summarize standardized differences in major imbalanced variables between the persistent residual albuminuria trajectory and the other two trajectories combined. Right-skewed variables, including UACR and UPCR, were log-transformed before analysis. Logistic regression was used to examine clinical correlates of the persistent residual albuminuria trajectory, with the favorable and partial trajectories combined as the reference. Because this trajectory included only 31 patients, the main model was limited to 4 variables selected before model fitting: baseline log-UACR, HbA1c, systolic blood pressure, and SGLT2 inhibitor use. eGFR and finerenone dose were considered only in an extended model. No a priori sample-size calculation was performed because all eligible patients during the study period were included. Missing covariates were handled by multiple imputation with chained equations when the missing-at-random assumption was considered plausible, with complete-case analysis used as a sensitivity analysis. Twenty imputed datasets were generated, and estimates were pooled using Rubin’s rules.22 Additional sensitivity analyses included stricter patient-level exclusion for potentially acute-event-affected follow-up, posterior-probability weighting, repeating the analysis using raw log-UACR level trajectories, and including patients with only 2 valid UACR time points in a supplemental 2-point change analysis. Week-24 eGFR change was compared across trajectory groups, and its association with log(UACR_week24/UACR_baseline) was examined using Spearman correlation and an exploratory linear model adjusted for baseline eGFR, baseline log-UACR, age, and sex. Because obesity may influence the relationship between spot UACR and daily albumin excretion, the correlate analysis was repeated with BMI adjustment and after excluding patients with BMI ≥28.5 kg/m2; directly measured UACR values were not converted to estimated 24-hour excretion.23 Use of Compound Centella Formula No. 2, TCM syndrome information, and inflammation- or fibrosis-related indicators remained descriptive only. All tests were 2-sided, and P values were interpreted together with effect sizes and 95% CIs. Statistical analyses were performed in R using lcmm, nnet, mice, and related packages.21,22 Results Patient Screening, Finerenone Treatment Timeline, and Baseline Characteristics During the study period, 218 patients with type 2 diabetes-related chronic kidney disease and a finerenone prescription or treatment assessment record were identified. Of these, 154 had residual albuminuria after at least 4 weeks of finerenone therapy and entered further screening. Nine patients were excluded for insufficient longitudinal UACR data, 5 for severe missingness of key clinical data, 3 for acute events affecting UACR interpretation, and 1 for another clearly defined kidney disease. The main trajectory analysis therefore included 136 patients, all with at least 3 valid UACR measurements within 24 weeks including index baseline (median, 4; interquartile range, 3–5) (Figure 1). Patients with only 2 valid UACR time points but otherwise complete data were used only in the supplemental 2-point change sensitivity analysis. Patient selection and trajectory classification flow diagram. Of 218 patients with type 2 diabetes-related CKD/DKD and a finerenone prescription or treatment assessment record, 154 entered further screening after at least 4 weeks of finerenone treatment with residual albuminuria. After exclusions for insufficient UACR time points, missing key clinical data, acute events affecting UACR interpretation, or other definite kidney disease, 136 patients with at least 3 valid UACR measurements within 24 weeks were included in the main trajectory analysis. Latent class mixed modeling identified 3 baseline-corrected UACR response trajectories: favorable response, partial response, and persistent residual albuminuria. Abbreviations: CKD, chronic kidney disease; DKD, diabetic kidney disease; LCMM, latent class mixed model; UACR, urinary albumin-to-creatinine ratio. The mean age was 61.5 ± 9.6 years, and 80 patients (58.8%) were male. At index baseline, median UACR was 704 mg/g (405, 1060), median UPCR was 1.22 g/g (0.75, 1.93), eGFR was 52.1 ± 13.0 mL/min/1.73 m2, HbA1c was 7.5 ± 1.1%, and systolic blood pressure was 135 ± 15 mmHg. The median interval from finerenone initiation to index baseline was 6.8 weeks (4.9, 10.2), and 38.2% received finerenone 20 mg once daily. ACEI/ARB, SGLT2 inhibitors, and GLP-1 receptor agonists were used by 88.2%, 62.5%, and 17.6%, respectively. BMI ≥28.5 kg/m2 was recorded in 26 patients (19.1%). Finerenone dose changes, background-treatment timing, and serum potassium monitoring are summarized in Supplementary Table S1; Baseline characteristics are shown in Table 1. Table 1Baseline Characteristics According to Baseline-Corrected UACR Response TrajectoriesCharacteristicOverall (n=136)Favorable (n=43)Partial (n=62)Persistent Residual (n=31)Age, years61.5 ± 9.660.4 ± 9.461.7 ± 9.562.8 ± 10.0Male sex80 (58.8)25 (58.1)37 (59.7)18 (58.1)Diabetes duration, years11.0 (7.0, 15.0)10.0 (6.5, 14.0)11.0 (7.0, 15.0)12.0 (8.0, 16.5)BMI, kg/m225.7 ± 3.325.4 ± 3.125.8 ± 3.326.1 ± 3.4BMI ≥28.5 kg/m226 (19.1)7 (16.3)12 (19.4)7 (22.6)Hypertension120 (88.2)36 (83.7)55 (88.7)29 (93.5)Cardiovascular disease43 (31.6)12 (27.9)20 (32.3)11 (35.5)Diabetic retinopathy67 (49.3)18 (41.9)30 (48.4)19 (61.3)UACR, mg/g704 (405, 1060)676 (385, 1010)714 (408, 1080)755 (445, 1160)UPCR, g/g1.22 (0.75, 1.93)1.12 (0.69, 1.78)1.25 (0.78, 1.95)1.33 (0.86, 2.10)eGFR, mL/min/1.73 m252.1 ± 13.053.4 ± 12.452.6 ± 13.049.2 ± 13.6HbA1c, %7.5 ± 1.17.2 ± 1.07.5 ± 1.17.9 ± 1.2Systolic BP, mmHg135 ± 15131 ± 14135 ± 15141 ± 15Diastolic BP, mmHg78 ± 1077 ± 978 ± 1080 ± 10Serum potassium, mmol/L4.46 ± 0.454.42 ± 0.424.47 ± 0.454.50 ± 0.48Hemoglobin, g/L123 ± 17125 ± 17123 ± 18121 ± 17Serum albumin, g/L38.1 ± 4.338.6 ± 4.138.1 ± 4.337.4 ± 4.4ALT, U/L22 (17, 31)21 (16, 30)22 (17, 31)24 (18, 34)ACEI/ARB use120 (88.2)38 (88.4)55 (88.7)27 (87.1)SGLT2 inhibitor use85 (62.5)32 (74.4)39 (62.9)14 (45.2)GLP-1 receptor agonist use24 (17.6)9 (20.9)11 (17.7)4 (12.9)Finerenone 20 mg once daily52 (38.2)15 (34.9)24 (38.7)13 (41.9)Finerenone-to-index interval, weeks6.8 (4.9, 10.2)6.9 (4.8, 10.1)6.6 (4.9, 9.8)7.1 (5.0, 10.7)Statin use100 (73.5)31 (72.1)45 (72.6)24 (77.4)Insulin use66 (48.5)18 (41.9)30 (48.4)18 (58.1)Metformin use88 (64.7)30 (69.8)40 (64.5)18 (58.1)Notes: Values are mean ± SD, median (IQR), or n (%). BMI ≥28.5 kg/m2 was used as the sensitivity-analysis threshold based on Chen et al and was not treated as a study-specific diagnostic definition of obesity.23.Abbreviations: ACEI, angiotensin-converting enzyme inhibitor; ALT, alanine aminotransferase; ARB, angiotensin receptor blocker; BMI, body mass index; BP, blood pressure; eGFR, estimated glomerular filtration rate; GLP-1, glucagon-like peptide-1; SGLT2, sodium-glucose cotransporter 2; UACR, urinary albumin-to-creatinine ratio; UPCR, urinary protein-to-creatinine ratio. Baseline Characteristics According to Baseline-Corrected UACR Response Trajectories Notes: Values are mean ± SD, median (IQR), or n (%). BMI ≥28.5 kg/m2 was used as the sensitivity-analysis threshold based on Chen et al and was not treated as a study-specific diagnostic definition of obesity.23. Abbreviations: ACEI, angiotensin-converting enzyme inhibitor; ALT, alanine aminotransferase; ARB, angiotensin receptor blocker; BMI, body mass index; BP, blood pressure; eGFR, estimated glomerular filtration rate; GLP-1, glucagon-like peptide-1; SGLT2, sodium-glucose cotransporter 2; UACR, urinary albumin-to-creatinine ratio; UPCR, urinary protein-to-creatinine ratio. Completeness of UACR Follow-Up Data and Identification of Response Trajectories The numbers of patients with available UACR data in the index baseline, week-4, week-8, week-12, and week-24 windows were 136, 112, 106, 119, and 128, respectively. The corresponding median actual testing times were 0.0, 4.3, 8.1, 12.4, and 24.2 weeks. The number of available samples in each window, actual testing times, and missingness of major variables are shown in Table 2 and Supplementary Table S2. Table 2LCMM Selection, Classification Quality, and UACR Follow-Up AvailabilityPanel A. Latent Class Mixed-Model SelectionModelAICBICEntropySmallest Class, %Mean Posterior Probability1-class model676.8699.5—100.0Reference model2-class model602.4649.00.7334.60.86 / 0.823-class model (selected)574.1641.10.7822.80.84 / 0.80 / 0.784-class model568.9663.30.747.40.83 / 0.77 / 0.72 / 0.70Panel B. Final Class Assignment and Classification QualityTrajectory classn (%)Mean PosteriorProbabilityMedian PosteriorProbability (IQR)Posterior Probability ≥0.70Favorable response43 (31.6)0.840.87 (0.79, 0.94)35/43 (81.4)Partial response62 (45.6)0.800.83 (0.75, 0.91)46/62 (74.2)Persistent residual albuminuria31 (22.8)0.780.81 (0.71, 0.89)22/31 (71.0)Panel C. UACR Follow-up-Window AvailabilityWindowAvailable n (%)Target WindowActual Timing,Median (IQR), WeeksBaseline136 (100.0)Week 00.0 (0.0, 0.0)Week 4112 (82.4)4 ± 2 weeks4.3 (3.4, 5.5)Week 8106 (77.9)8 ± 2 weeks8.1 (7.1, 9.5)Week 12119 (87.5)12 ± 2 weeks12.4 (11.1, 13.6)Week 24128 (94.1)24 ± 2 weeks24.2 (23.1, 25.5)Notes: The final 3-class model used a linear time term and an individual-level random intercept and was selected based on model fit, classification quality, class size, convergence, and clinical interpretability.Abbreviations: AIC, Akaike information criterion; BIC, Bayesian information criterion; IQR, interquartile range; LCMM, latent class mixed model; UACR, urinary albumin-to-creatinine ratio. LCMM Selection, Classification Quality, and UACR Follow-Up Availability Notes: The final 3-class model used a linear time term and an individual-level random intercept and was selected based on model fit, classification quality, class size, convergence, and clinical interpretability. Abbreviations: AIC, Akaike information criterion; BIC, Bayesian information criterion; IQR, interquartile range; LCMM, latent class mixed model; UACR, urinary albumin-to-creatinine ratio. Based on longitudinal log (UACR_t/UACR_baseline) data over 24 weeks, the latent class mixed model identified 3 baseline-corrected UACR response trajectories. Compared with the 2-class model, the 3-class model showed better model fit and clinical interpretability. Although the 4-class model had a slightly lower AIC, it had a higher BIC, a smallest class proportion of only 7.4%, and insufficient stability of the classification structure. It was therefore not selected as the final model. The final model used a linear time term and an individual-level random intercept. Alternative specifications with a quadratic time term or random slope did not further improve model stability or clinical interpretability. The 3 trajectories were a favorable response trajectory in 43 patients (31.6%), a partial response trajectory in 62 patients (45.6%), and a persistent residual albuminuria trajectory in 31 patients (22.8%). The average posterior classification probabilities for the 3 classes were 0.84, 0.80, and 0.78, respectively. Model fit indices, classification quality, and follow-up window availability are shown in Table 2. Among patients with available UACR data in the week-24 window, the median relative changes in UACR were −51.6%, −24.8%, and −4.9% in the favorable response, partial response, and persistent residual albuminuria trajectories, respectively. Partial response or persistent residual albuminuria accounted for 93 of 136 patients (68.4%). Figure 2 summarizes the 3 baseline-corrected UACR response patterns, week-24 response thresholds, and classification quality. Baseline-corrected UACR response trajectories, 24-week response thresholds, and classification quality. (A) provides a visual summary of baseline-corrected UACR response patterns using log(uacr_t/UACR_baseline), anchored at index baseline and the reported median relative changes at week 24. UACR sample sizes at baseline and weeks 4, 8, 12, and 24 were 136, 112, 106, 119, and 128, respectively. (B) shows the proportions achieving ≥30% and ≥50% UACR reduction among patients with available week-24 data. (C) summarizes classification quality for the selected 3-class LCMM. Abbreviations: APPA, average posterior probability of assignment; LCMM, latent class mixed model; UACR, urinary albumin-to-creatinine ratio. Clinical Correlates of the Persistent Residual Albuminuria Trajectory Compared with the other 2 trajectories, patients in the persistent residual albuminuria trajectory had relatively higher HbA1c and systolic blood pressure at the index baseline and a lower proportion of SGLT2 inhibitor use. eGFR, diabetic retinopathy, and baseline UACR also showed directional differences. Complete baseline characteristics are shown in Table 1. In the multivariable logistic model with the persistent residual albuminuria trajectory as the outcome and the other 2 trajectories combined as the reference group, the adjusted OR for each 1% higher HbA1c was 1.46 (95% CI, 1.07–2.00). The adjusted OR for each 10-mmHg higher systolic blood pressure was 1.28 (95% CI, 1.01–1.63), and the adjusted OR for SGLT2 inhibitor use was 0.50 (95% CI, 0.25–0.99). Baseline log-UACR showed a consistent but imprecise direction of association, with an adjusted OR of 1.38 (95% CI, 0.88–2.16). After eGFR and finerenone dose were added to the extended model, the confidence intervals for systolic blood pressure and SGLT2 inhibitor use crossed the null value, indicating remaining uncertainty. Results of the correlate analysis are shown in Table 3 and Figure 3. Table 3Clinical Correlates and Key 24-Week Outcomes According to UACR Response TrajectoryPanel A. Clinical Correlates of the Persistent Residual Albuminuria TrajectoryVariableMain ModelOR (95% CI)P valueExtended ModelOR (95% CI)InterpretationBaseline log-UACR, per 1-unit increase1.38 (0.88–2.16)0.161.34 (0.85–2.13)Directionally consistent but impreciseHbA1c, per 1% increase1.46 (1.07–2.00)0.0181.43 (1.04–1.97)Higher estimated oddsSystolic BP, per 10-mmHg increase1.28 (1.01–1.63)0.0441.25 (0.98–1.61)Higher estimated odds; imprecise in extended modelSGLT2 inhibitor use0.50 (0.25–0.99)0.0470.52 (0.26–1.05)Lower estimated odds; imprecise in extended modeleGFR, per 10 mL/min/1.73 m2 increaseNot included—0.90 (0.68–1.20)Exploratory extensionFinerenone 20 mg once dailyNot included—1.12 (0.54–2.32)Exploratory extensionPanel B. Key 24-Week Clinical OutcomesOutcomeOverall (n=136)Favorable (n=43)Partial (n=62)Persistent Residual (n=31)UACR reduction ≥30%, n/N (%)61/128 (47.7)31/41 (75.6)25/58 (43.1)5/29 (17.2)UACR reduction ≥50%, n/N (%)33/128 (25.8)21/41 (51.2)10/58 (17.2)2/29 (6.9)Median relative UACR change at week 24, %−25.9−51.6−24.8−4.9Relative UPCR change at week 24, %, n=124−19.6 (−40.5, 3.5)−41.0 (−58.5, −18.0)−20.5 (−38.0, 2.0)−3.8 (−21.0, 21.5)eGFR change at week 24, mL/min/1.73 m2, n=136−1.3 ± 5.9; P=0.60−0.8 ± 5.6 (−2.5 to 0.9)−1.3 ± 5.9 (−2.8 to 0.2)−2.2 ± 6.3 (−4.5 to 0.1)HbA1c change at week 24, %, n=132−0.18 ± 0.80−0.25 ± 0.75−0.18 ± 0.80−0.12 ± 0.85Systolic BP change at week 24, mmHg, n=136−4 ± 15−5 ± 14−4 ± 15−3 ± 15Notes: The main model included baseline log-UACR, HbA1c, systolic BP, and SGLT2 inhibitor use. The extended model additionally considered eGFR and finerenone dose. Continuous secondary outcomes are descriptive; eGFR values in parentheses are 95% CIs, and P=0.60 is the overall trajectory-group comparison.Abbreviations: CI, confidence interval; OR, odds ratio. Clinical Correlates and Key 24-Week Outcomes According to UACR Response Trajectory Notes: The main model included baseline log-UACR, HbA1c, systolic BP, and SGLT2 inhibitor use. The extended model additionally considered eGFR and finerenone dose. Continuous secondary outcomes are descriptive; eGFR values in parentheses are 95% CIs, and P=0.60 is the overall trajectory-group comparison. Abbreviations: CI, confidence interval; OR, odds ratio. Clinical correlates of the persistent residual albuminuria trajectory. Odds ratios compare the persistent residual albuminuria trajectory with the favorable and partial response trajectories combined. Filled circles represent the main model, and open circles represent the extended model. The dashed vertical line indicates OR=1. The main model included baseline log-UACR, HbA1c, systolic blood pressure, and SGLT2 inhibitor use; the extended model additionally considered eGFR and finerenone dose. Values are observational associations and should not be interpreted causally. Abbreviations:CI, confidence interval; eGFR, estimated glomerular filtration rate; OR, odds ratio; SGLT2, sodium-glucose cotransporter 2; UACR, urinary albumin-to-creatinine ratio. Week-24 Outcomes and Safety Events Among patients with available week-24 UACR data, the proportions achieving a UACR reduction of at least 30% were 75.6%, 43.1%, and 17.2% in the favorable response, partial response, and persistent residual albuminuria trajectories, respectively. The corresponding proportions achieving a UACR reduction of at least 50% were 51.2%, 17.2%, and 6.9%. UPCR changes were broadly consistent with UACR changes, although individual variability was greater. Mean week-24 eGFR changes were −0.8, −1.3, and −2.2 mL/min/1.73 m2, respectively, without a clear overall difference (P=0.60). Among 128 patients with paired week-24 UACR and eGFR data, log(UACR_week24/UACR_baseline) was weakly correlated with eGFR change (Spearman ρ=-0.10; P=0.27); the adjusted linear estimate was likewise imprecise (β=-0.84 mL/min/1.73 m2 per log-unit; 95% CI, −2.58 to 0.90; P=0.34). HbA1c and systolic blood pressure decreased slightly in all 3 trajectories. Major outcomes are shown in Table 3 and Supplementary Table S2. The safety analysis included all 136 patients. Among 132 patients with follow-up serum potassium measurements, hyperkalemia occurred in 18 patients (13.6%), including 3 patients (2.3%) with serum potassium of at least 6.0 mmol/L. Among patients with the relevant follow-up tests, ALT or AST elevation greater than 3 times the upper limit of normal occurred in 4 of 133 patients (3.0%), leukopenia in 3 of 132 patients (2.3%), thrombocytopenia in 1 of 132 patients (0.8%), and hemoglobin decrease greater than 20 g/L in 7 of 132 patients (5.3%). Gastrointestinal discomfort was recorded in 13 of 136 patients (9.6%), and all-cause hospitalization in 11 of 136 patients (8.1%). No death, dialysis initiation, or definite treatment-related serious adverse event was recorded within 24 weeks. The overall incidence of safety events was low, and these results should be interpreted as descriptive safety findings only (Supplementary Table S1). Sensitivity Analyses and Exploratory Variables Sensitivity analyses generally supported the 3-trajectory structure, although precision varied. Under the stricter patient-level acute-event exclusion, agreement with the main classification was 87.7% (weighted κ=0.81). In posterior-probability weighted, complete-case, and multiple-imputation analyses, the HbA1c estimate remained directionally similar, whereas systolic blood pressure and SGLT2 inhibitor estimates were less stable. The supplemental 2-point analysis included 9 additional patients (total n=145) and was used only to assess directional consistency. Raw log-UACR trajectory agreement was 72.1% (weighted κ=0.58). Adding BMI to the correlate model did not materially alter the HbA1c estimate (OR, 1.44; 95% CI, 1.05–1.98), and excluding patients with BMI ≥28.5 kg/m2 produced a similar but less precise estimate (OR, 1.42; 95% CI, 1.01–2.00; n=110). First-prescription dates were traceable for 110 of 120 ACEI/ARB users and 74 of 85 SGLT2 inhibitor users at index baseline. Descriptive week-24 UACR changes varied across treatment-timing categories, with numerically greater reductions among patients who had initiated SGLT2 inhibitors before finerenone than among those who initiated them later or never used them. Given the small, uneven subgroups and nonrandom treatment allocation, no formal hypothesis testing or adjusted modeling was performed (Supplementary Table S1). Records of Compound Centella Formula No. 2 use, TCM syndrome scores, and inflammation- or fibrosis-related indicators were described only as exploratory real-world variables. Because the number of tested patients was limited, missingness was substantial, and no stable mechanistic evidence emerged, these results were not used to infer the efficacy of Chinese herbal medicine, TCM syndrome mechanisms, or the role of inflammatory or fibrotic pathways. The availability and directional summaries of exploratory variables are shown in Supplementary Table S2. Discussion Main Findings Using single-center real-world longitudinal data and the baseline-corrected log(UACR_t/UACR_baseline) ratio, this study identified 24-week UACR response trajectories in patients with DKD and residual albuminuria during finerenone therapy. The results showed that albuminuria reduction in this population was not uniform or homogeneous. Instead, patients could be classified into 3 trajectories: favorable response, partial response, and persistent residual albuminuria. Approximately one-third of patients showed a marked and relatively sustained reduction in UACR, nearly half showed a moderate reduction, and approximately one-quarter had limited UACR reduction or persistent fluctuating residual albuminuria over 24 weeks. These findings suggest that short-term albuminuria responses remain heterogeneous even during finerenone and individualized kidney-protective therapy. Partial response or persistent residual albuminuria accounted for 68.4% of this selected cohort, underscoring that most patients continued to show incomplete short-term UACR reduction. Compared with overall change or a single binary threshold, a multi-time-point baseline-corrected trajectory approach can describe when and to what extent responses diverge.15,16 This study describes routine-practice response patterns and correlates rather than causal treatment effects. Clinical Implications: From Single Time-Point UACR to Dynamic Risk Identification UACR is an important indicator for DKD risk stratification and treatment-response assessment. However, UACR is influenced by glycemia, blood pressure, infection, acute kidney-function fluctuation, sampling time, and adjustment of background therapy, resulting in substantial within- and between-patient variability.4,5 Finerenone has become an important component of modern kidney-protective therapy for chronic kidney disease associated with type 2 diabetes. However, some patients in routine practice continue to have residual albuminuria after treatment.10,12,13 The persistent residual albuminuria trajectory identified in this study suggests that these patients have limited albuminuria reduction over 24 weeks and may require closer dynamic follow-up and more comprehensive risk-factor management. Trajectory classification is not intended to replace established treatment-response criteria. It may instead support closer reassessment of adherence, blood pressure and glycemic control, salt intake, background kidney-protective treatment, finerenone dose, and hyperkalemia risk in patients with limited short-term response.2,20,24 The trajectories remain statistical summaries rather than validated prognostic subtypes, and their use for treatment decisions requires external validation.3,13 Clinical Correlates and Interpretive Boundaries This study observed that higher HbA1c and higher systolic blood pressure were associated with higher estimated odds of the persistent residual albuminuria trajectory, whereas SGLT2 inhibitor use showed a directionally consistent association with lower estimated odds of this trajectory. Baseline log-UACR showed a consistent direction of association, but the estimate was imprecise in the baseline-corrected trajectory model. Because the main trajectory model used UACR change relative to each patient’s own baseline rather than raw UACR levels, the identified trajectories emphasize dynamic response patterns and should not be interpreted simply as strata of baseline albuminuria burden. These findings are broadly consistent with current understanding of DKD progression and persistent albuminuria. Poor glycemic control and higher systolic blood pressure may affect albuminuria through microvascular injury, glomerular filtration barrier dysfunction, and hemodynamic load.2,20,24 The directionally lower estimated odds of the persistent residual albuminuria trajectory among SGLT2 inhibitor users is also consistent with the clinical rationale for multi-mechanistic kidney-protective therapy.7,8,25 However, these factors should be interpreted only as clinical correlates of trajectory class, not as causal determinants. Unmeasured factors, including treatment adherence, salt intake, home blood pressure, background therapy adjustment, and follow-up frequency, may all influence UACR changes. The estimate related to SGLT2 inhibitor use also remained uncertain and requires validation in larger samples. Week-24 Outcomes, Safety, and Exploratory Variables The proportions achieving UACR reductions of at least 30% and at least 50% were highest in the favorable trajectory and lowest in the persistent trajectory. Because both the trajectories and thresholds derive from UACR change, these results support clinical interpretability but are not independent validation. Neither trajectory-group comparisons nor continuous analyses showed a clear association between UACR response and 24-week eGFR change. This does not establish long-term benefit or harm because 24 weeks is insufficient to evaluate sustained eGFR slope, kidney failure, cardiovascular events, or death.1,13,26 Recent evidence also suggests that spot UACR may underestimate daily albumin excretion in some patients with obesity. We analyzed directly measured UACR relative to each patient’s own baseline rather than estimating 24-hour excretion; BMI-adjusted and BMI-restricted analyses did not materially change the principal correlate finding, although residual measurement bias cannot be excluded.23 Regarding safety, no death, dialysis initiation, or definite treatment-related serious adverse event was recorded within 24 weeks. The incidence of hyperkalemia was 13.6%, and the incidence of severe hyperkalemia was 2.3%. Safety events were summarized descriptively and were not formally compared across trajectory groups. Elevations in liver enzymes, leukopenia, thrombocytopenia, and gastrointestinal discomfort were infrequent. However, this study provided only descriptive safety analyses and cannot establish long-term safety.27,28 Use of Compound Centella Formula No. 2, TCM syndrome information, and inflammation- or fibrosis-related indicators was described only as exploratory variables. These data were not used to evaluate the efficacy of Chinese herbal medicine, verify TCM syndrome mechanisms, or demonstrate the role of inflammatory or fibrotic pathways. These variables are susceptible to treatment selection, disease severity, physician prescribing preference, patient adherence, testing accessibility, and record completeness. They should therefore be considered only as reference signals for future study design. Methodological Considerations and Limitations This study used baseline-corrected log-UACR ratios to reduce the extent to which trajectory classification reflected baseline albuminuria burden. The limited agreement with raw log-UACR trajectories suggests that the 2 approaches capture related but different information. Visit-specific availability, actual testing times, classification quality, and multiple sensitivity analyses were reported.19,21 Because all eligible patients during the study period were included, no a priori sample-size calculation was performed; the small persistent trajectory nevertheless limits multivariable precision and increases overfitting risk. This study has several limitations. It was a single-center retrospective cohort with a limited sample size, especially in the persistent residual albuminuria trajectory, and residual confounding, imprecision, and overfitting remain possible. UACR has substantial biological and analytical variability; measurement fluctuation, regression to the mean, obesity-related creatinine effects, and unrecorded acute factors cannot be fully excluded.4,5,23 Requiring at least 3 UACR measurements may have introduced selection bias related to follow-up completeness. Treatment timing could not be ascertained for every patient, and the small, nonrandom timing subgroups precluded causal sequencing comparisons. LCMM class assignment is model-dependent and should not be interpreted as a stable biological subtype.19,21 Finally, 24 weeks is insufficient to evaluate long-term eGFR slope or hard kidney and cardiovascular outcomes. Conclusion In this single-center retrospective 24-week real-world study, 3 baseline-corrected UACR response trajectories were identified among patients with DKD and residual albuminuria during finerenone therapy. Partial or persistent response patterns accounted for 68.4% of this selected cohort. Higher HbA1c was associated with higher estimated odds of the persistent residual albuminuria trajectory, whereas other clinical-factor and short-term eGFR estimates remained uncertain. The findings are exploratory and require validation in larger, prospective, multicenter cohorts with longer follow-up before trajectory analysis can be used for clinical decision-making. Funding Statement This work was supported by the Wenzhou Municipal Science and Technology Project (Y2023205), the Zhejiang Provincial TCM Science and Technology Project (2023ZL676), and the Zhejiang Chinese Medical University Research Project (2022FSYYZY30). The funders had no role in study design, data collection, data analysis, interpretation of results, paper preparation, or the decision to submit the paper for publication. Sustainability Statement This retrospective study used existing electronic medical records and did not require additional patient visits, sample collection, or laboratory procedures. Data extraction and analysis were conducted electronically, thereby minimizing additional travel, consumable use, and research-related environmental burden. Data Sharing Statement Data sharing is subject to institutional data-governance and ethical requirements. Requests may be directed to the corresponding author. Ethics Approval and Consent to Participate This study was approved by the Ethics Committee of Wenzhou Hospital of Integrated Traditional Chinese and Western Medicine affiliated to Zhejiang Chinese Medical University (approval No. 2023-K049). The study was conducted in accordance with the Declaration of Helsinki and its later amendments. Because this was a retrospective study using de-identified routine clinical data and involved no additional patient intervention, the requirement for written informed consent was waived by the Ethics Committee. Author Contributions Yang Zhang: Conceptualization, Methodology, Formal analysis, Writing - original draft, Writing - review & editing. Qin Zhu, Chaochao Wang, Xiaoqiao Cai, Shengfen Lin, Beibei Ding, and Xingyu Zhu: Investigation, Data curation, Validation, Writing - review & editing. Yongqiang Lin: Conceptualization, Methodology, Supervision, Project administration, Writing - review & editing. All authors gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work. Disclosure The authors declare that they have no competing interests. References Associated Data Data Availability Statement Data sharing is subject to institutional data-governance and ethical requirements. Requests may be directed to the corresponding author.