Introduction: The Need for Precision in Diabetic Kidney Disease
Chronic kidney disease (CKD) in the context of type 2 diabetes (T2D) remains a significant global health burden. Despite the transformative impact of sodium-glucose cotransporter 2 (SGLT2) inhibitors, clinicians still face challenges in identifying which patients are at the highest risk for rapid progression and which are deriving the most benefit from therapy. Traditional markers, such as the estimated glomerular filtration rate (eGFR) and the urinary albumin-to-creatinine ratio (UACR), while foundational, often fail to capture the underlying molecular pathophysiology of kidney injury and inflammation.
This gap in prognostic and monitoring capabilities has led to the development of kidneyintelX.dkd, a biomarker-informed risk score that integrates tumor necrosis factor receptor-1 (TNFR-1), TNFR-2, and kidney injury molecule-1 (KIM-1) with clinical data. A recent analysis of the CANVAS and CREDENCE trials provides robust evidence that this score offers clinical utility far beyond traditional Kidney Disease Improving Global Outcomes (KDIGO) risk classifications.
Study Design and Methodology
The study utilized banked plasma samples from two landmark randomized controlled trials: the Canagliflozin Cardiovascular Assessment Study (CANVAS) and the Canagliflozin and Renal Events in Diabetes with Established Nephropathy Clinical Evaluation (CREDENCE). The researchers evaluated 2,954 participants who had CKD stages G1 to G3b.
The primary objective was to assess the prognostic utility of kidneyintelX.dkd at baseline and its longitudinal changes at one year. The composite kidney outcome included a sustained 40% decline in eGFR, kidney failure (dialysis, transplantation, or sustained eGFR <15 mL/min/1.73 m2), or death due to kidney-related causes. The study also examined how the score responded to canagliflozin treatment compared to placebo and whether those changes correlated with future clinical outcomes.
Key Findings: Superior Risk Stratification
At baseline, the kidneyintelX.dkd score categorized participants into three risk levels: low (26.0%), moderate (44.7%), and high (29.4%). The results demonstrated a clear log-linear association between the score and kidney outcomes. Specifically, each doubling of the kidneyintelX.dkd score at baseline was associated with a 2.20-fold increase in the risk of the composite kidney outcome (95% CI 1.72–2.82).
Table 1. Baseline characteristics of the total population and subgroups defined by kidneyintelX.dkd risk category
| Characteristic | Total | Low risk | Moderate risk | High risk |
|---|---|---|---|---|
| Participants, n | 2,954 | 767 | 1,320 | 867 |
| KidneyintelX.dkd | 0.17 [0.10; 0.34] | 0.09 [0.09; 0.10] | 0.16 [0.12; 0.21] | 0.44 [0.36; 0.50] |
| HbA1c, % | 8.24 (1.19) | 8.03 (0.85) | 8.27 (1.23) | 8.38 (1.34) |
| Systolic blood pressure, mmHg | 140 (16.0) | 138 (15.5) | 139 (15.6) | 143 (16.5) |
| BMI, kg/m2 | 32.1 (6.05) | 32.6 (6.18) | 32.0 (5.85) | 31.8 (6.22) |
| History of CVD, yes | 1,616 (54.7) | 431 (56.2) | 717 (54.3) | 468 (54.0) |
| RAAS inhibition, yes | 848 (85.0) | 505 (84.6) | 300 (86.7) | 43 (78.2) |
| UACR, mg/g | 518 [124; 1,231] | 57.4 [26.4; 161] | 502 [255; 786] | 1,718 [1,106; 2,710] |
| Normoalbuminuria (<30 mg/g) | 268 (9.07) | 198 (25.8) | 69 (5.23) | 1 (0.12) |
| Microalbuminuria (30–300 mg/g) | 806 (27.3) | 484 (63.1) | 312 (23.6) | 10 (1.15) |
| Macroalbuminuria (>300 mg/g) | 1,880 (63.6) | 85 (11.1) | 939 (71.1) | 856 (98.7) |
| eGFR, mL/min/1.73 m2 | 61.8 (19.1) | 70.1 (19.1) | 62.0 (18.1) | 54.2 (17.3) |
| <60 mL/min/1.73 m2 | 1,514 (51.3) | 300 (39.1) | 643 (48.7) | 571 (65.9) |
| ≥60 mL/min/1.73 m2 | 1,440 (48.7) | 467 (60.9) | 677 (51.3) | 296 (34.1) |
| KDIGO risk | ||||
| Low | 808 (27.4) | 590 (76.9) | 218 (16.5) | 0 (0.00) |
| Moderate | 1,050 (35.5) | 134 (17.5) | 617 (46.7) | 299 (34.5) |
| High | 1,096 (37.1) | 43 (5.61) | 485 (36.7) | 568 (65.5) |
| KIM-1, pg/mL | 231 [133; 420] | 108 [79.0; 144] | 226 [162; 311] | 555 [385; 820] |
| TNFR1, pg/mL | 3,662 [2,826; 4,783] | 2,816 [2,334; 3,442] | 3,639 [2,913; 4,596] | 4,845 [3,827; 5,982] |
| TNFR2, pg/mL | 13,836 [10,632; 18,157] | 10,634 [8,555; 13,035] | 13,643 [10,781; 17,689] | 18,095 [14,524; 22,348] |
When compared directly to the established KDIGO risk categories, kidneyintelX.dkd demonstrated superior precision. A categorical net reclassification analysis showed a total net reclassification index (NRI) of 21.5%. This improvement was driven by correctly identifying higher risk in patients who eventually experienced events and correctly identifying lower risk in those who did not. In practical terms, this corresponds to approximately 78 additional patients per 1,000 being correctly reclassified compared to using KDIGO criteria alone.
Fig 1. Association between baseline kidneyintelX.dkd and composite kidney outcome analyzed on a continuous scale. Dotted lines represent median kidneyintelX.dkd for each KDIGO risk classification (from left to right: low, moderate, high).
Fig 2. Risk for composite kidney outcome stratified by kidneyintelX.dkd across KDIGO risk strata and joint kidneyintelX.dkd and KDIGO risk classification. A: Association is shown between baseline kidneyintelX.dkd risk strata and the composite kidney outcome (40% eGFR decline, kidney failure, or renal death), across baseline KDIGO risk categories. B: Association is shown three-dimensionally between baseline kidneyintelX.dkd and KDIGO risk strata and the composite kidney outcome.
Fig 3. Association between change in kidneyintelX.dkd from baseline to year 1 and composite kidney outcome analyzed on a continuous scale and the distribution of changes by treatment group. Dotted lines represent changes of −25%,−10%, 10%, and 25%, respectively. Association is shown continuously between the change in kidneyintelX.dkd and the composite kidney outcome (40% eGFR decline, kidney failure, or renal death), adjusted for log-transformed kidneyintelX.dkd at baseline.
Longitudinal Monitoring and Treatment Response
One of the most clinically significant findings of this study is the responsiveness of kidneyintelX.dkd to therapeutic intervention. At the one-year mark, participants treated with canagliflozin showed a significant reduction in their kidneyintelX.dkd scores compared to those on placebo.
Furthermore, these longitudinal changes were predictive of future risk. A decrease in the kidneyintelX.dkd score by one standard deviation at year one was associated with a significant reduction in the subsequent risk of the composite kidney outcome (HR 1.56; 95% CI 1.28–1.90). Crucially, this association remained significant even after adjusting for changes in eGFR and UACR, suggesting that the biomarker score captures therapeutic responses that are not visible through traditional clinical metrics.
Individual Biomarker Performance
The study also dissected the individual contributions of the three biomarkers. At baseline, the hazard ratios per standard deviation increase were:
TNFR-1: 1.80 (95% CI 1.51–2.14)
TNFR-2: 1.71 (95% CI 1.45–2.02)
KIM-1: 1.93 (95% CI 1.67–2.23)
While individual biomarkers were strong predictors, the integrated kidneyintelX.dkd score consistently showed slightly stronger associations with kidney outcomes, highlighting the value of a multi-marker approach that accounts for both inflammation (TNFR-1/2) and tubular injury (KIM-1).
Expert Commentary: Mechanistic Insights and Clinical Implications
The ability of kidneyintelX.dkd to outperform KDIGO classification stems from its focus on active biological processes. TNFR-1 and TNFR-2 are markers of systemic and local inflammation, while KIM-1 is a highly specific indicator of proximal tubular damage. In diabetic kidney disease, these processes often precede the structural damage that leads to a drop in eGFR or the barrier dysfunction that results in albuminuria.
From a clinical perspective, these findings suggest that kidneyintelX.dkd could serve as a “liquid biopsy,” providing a real-time window into kidney health. The observation that canagliflozin reduces these markers suggests that SGLT2 inhibitors do more than just alter hemodynamics; they actively mitigate the inflammatory and injurious pathways that drive disease progression. For clinicians, the high absolute risk reduction seen in the “high-risk” kidneyintelX.dkd group confirms that this score can identify the patients who stand to gain the most from intensive SGLT2 inhibitor therapy.
Study Limitations
While the results are compelling, the researchers noted some limitations. The analysis was post-hoc and conducted within the context of controlled clinical trials, which may not perfectly reflect real-world diversity. Additionally, the cohort excluded patients in the lowest KDIGO risk categories and those with very advanced CKD (G4-G5), meaning the findings are most applicable to the early-to-mid stages of disease progression.
Conclusion: A New Era of Personalized Nephrology
The validation of kidneyintelX.dkd in the CANVAS and CREDENCE cohorts marks a significant step toward personalized medicine in nephrology. By providing a more nuanced risk profile and a responsive tool for monitoring treatment, this score allows for earlier intervention and more precise management of patients with type 2 diabetes and CKD. As we move forward, integrating such biomarker-based assessments into routine clinical practice could help reduce the incidence of kidney failure and improve the long-term prognosis for millions of patients worldwide.
References
Moedt E, Coca SG, Edwards K, Neuen BL, Arnott C, Bakker SJL, Fleming F, Heerspink HJL. Baseline Risk and Longitudinal Changes in kidneyintelX.dkd and Its Association With Kidney Outcomes in the CANVAS and CREDENCE Trials. Diabetes Care. 2026 Jan 1;49(1):92-98. doi: 10.2337/dc25-1722 IF: 16.6 Q1 . PMID: 41217780 IF: 16.6 Q1 .
