Time‐Dependent Accrual of Cardiorenal Benefit With SGLT2 Inhibitors in Chronic Kidney Disease: A Reconstructed Individual Patient Data Meta‐Analysis of Randomized Trials

Abbreviations

CKD

chronic kidney disease

eGFR

estimated glomerular filtration rate

RMST

restricted mean survival time

SGLT2i

Sodium‐glucose cotransporter 2 inhibitors

1. Background

Sodium‐glucose cotransporter 2 inhibitors (SGLT2i) are recommended for patients with chronic kidney disease (CKD) and an estimated glomerular filtration rate (eGFR) ≥ 20 mL/min/1.73 m² with albuminuria [1]. Although SGLT2i induce an early decline in eGFR [2], this initial dip has not been associated with adverse clinical outcomes; current guidelines and reviews do not recommend discontinuation of therapy in the absence of substantial kidney function decline [1, 3].

Nevertheless, treatment discontinuation remains common in routine clinical practice [4, 5, 6]. In a real‐world cohort of patients with advanced CKD, discontinuation rates reached 10% and 17% at 6 and 12 months of follow‐up, respectively [5]. Of these, two‐thirds of the discontinuations were attributed to declines in eGFR. In a contemporary HF registry, the median time to treatment discontinuation was 4.9 months, and poor kidney function was associated with an increased risk of discontinuation [6]. These observations highlight a persistent gap between guideline recommendations and real‐world clinical practice.

Previous studies have primarily focused on demonstrating the overall efficacy of SGLT2i therapy in reducing cumulative clinical events. However, the temporal pattern of clinical benefit accumulation has not been systematically investigated. In this study, we pooled placebo‐controlled randomized controlled trials (RCTs) of SGLT2i therapy in patients with CKD. By evaluating the temporal evolution of restricted mean survival time (RMST) differences for cardiorenal outcomes [7], we sought to quantitatively characterize how the clinical benefit of SGLT2i accumulates over time.

2. Methods

Following PRISMA guidelines, we systematically searched PubMed, Scopus and the Cochrane Library in the middle of June 2026. The complete search strategies for all databases are provided in Table S1. Institutional review board approval was not required because only published data were used. RCTs comparing SGLT2i with placebo in patients with CKD were considered eligible for inclusion. Two authors (S.Y. and M.I.) independently screened studies for eligibility. Time‐to‐event individual patient data (IPD) were reconstructed from published Kaplan–Meier curves using a previously validated algorithm [8]. The reconstructed IPD were generated using the IPDfromKM package in R, incorporating digitized coordinates, numbers at risk and reported event counts. A two‐stage IPD meta‐analysis was conducted. In the first stage, Kaplan–Meier survival functions were estimated separately for each treatment group. At each time point t, the RMST up to t was calculated and the RMST difference was obtained within each trial, following a previously described framework [9]. The variance of the RMST difference at each time point was estimated using 5000 bootstrap resampling within each trial. In the second stage, trial‐specific RMST differences were pooled using fixed‐effect inverse‐variance weighting under a normal approximation. Corresponding 95% confidence intervals (CI) were calculated analytically. In addition, heterogeneity at months 10, 20 and 30 was assessed using Cochran's Q and I ². For primary analysis, follow‐up time was truncated at 30 months for all trials to ensure temporal alignment and comparability. An exploratory analysis including the full follow‐up of CREDENCE was also performed. As a conventional summary measure, pooled HRs were estimated using a stratified Cox proportional hazards model with trial as the stratification factor, using follow‐up truncated at τ. Heterogeneity was assessed using a treatment‐by‐trial interaction and Cochran's Q and I ² statistics.

Risk of bias was assessed using the Cochrane Risk of Bias 2 (RoB 2) tool, and the certainty of evidence was evaluated using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework for the primary outcomes. Both assessments were performed independently by two authors, with disagreements resolved by consensus.

3. Results

The search identified 1977 records. After title and abstract screening, 19 studies were assessed in full text and three RCTs (CREDENCE [10], DAPA‐CKD [11], EMPA‐KIDNEY [12] trial) met the eligibility criteria (Figure S1). A total of 15 314 patients were analysed (Table 1). The reconstructed IPD showed good agreement with the original data (Tables S2 and S3; Figure S2).

TABLE 1.

Characteristics of the studies included in the quantitative synthesis.

Trial	Year	Follow‐up (months)	N (SGLT2i/placebo)	Mean age	Male sex	Diabetes (%)	BMI	CV disease (%)	HF (%)	Primary outcome
CREDENCE [10]	2019	42	2202/2199	63.0	66.1	100	31.3	50.5	14.8	ESKD (dialysis for at least 30 days, kidney transplantation, or an eGFR of < 15 mL/min/1.73 m2 sustained ≥ 30 days) Doubling of the serum creatinine level from baseline sustained ≥ 30 days Death from renal or CV disease
DAPA‐CKD [11]	2020	32	2152/2152	61.85	66.9	67.5	29.5	37.4	10.9	Decline of at least 50% in the eGFR (confirmed after ≥ 28 days) ESKD (maintenance dialysis for ≥ 28 days, kidney transplantation, or an eGFR of < 15 mL/min/1.73 m2 confirmed after ≥ 28 days) Death from renal or CV causes
EMPA‐KIDNEY [12]	2023	30	3304/3305	63.9	66.8	46.0	29.8	26.7	10.0	ESKD (maintenance dialysis or receipt of a kidney transplant) Sustained decrease in the eGFR to less than 10 mL/min/1.73 m2 Sustained decrease from baseline in the eGFR of at least 40% Death from renal or CV causes

Abbreviations: BMI, body mass index; CV, cardiovascular; eGFR, estimated glomerular filtration rate; ESKD, end‐stage kidney disease; SGLT2i, sodium‐glucose cotransporter 2 inhibitors.

SGLT2i was associated with a significantly lower risk of cardiorenal outcome (HR 0.68; 95% CI, 0.62–0.74; p < 0.001). Trial‐specific analyses demonstrated no evidence of between‐trial heterogeneity (I ² = 12.8%; Cochran Q p = 0.32). The proportional hazards assumption was assessed using Schoenfeld residuals, and no significant violation was detected (p = 0.14).

The RMST difference was 0.36 days (95% CI, −0.31 to 1.04) at 10 months, 5.01 days (95% CI, 2.67–7.35) at 20 months and 17.0 days (95% CI, 12.0–22.0) at 30 months (Figure 1). Between‐trial heterogeneity remained low (I ² = 6.8%, 0% and 32.7% at 10, 20 and 30 months, respectively). In an exploratory analysis including the full follow‐up of CREDENCE, the RMST difference reached 40.0 days (95% CI, 30.5–49.5) at 42 months (Figure S3). In the secondary analysis of cardiovascular death, DAPA‐CKD was excluded because Kaplan–Meier curves for cardiovascular death were not reported. The RMST difference was 0.04, 0.59, 1.78 and 4.70 days at 10, 20, 30 and 42 months, respectively (Figure S4). These findings should be interpreted cautiously because the analysis included only two trials.

FIGURE 1.

Pooled RMST difference curve for cardiorenal composite endpoint. Follow‐up time was truncated at 30 months for all trials to ensure temporal alignment and comparability. p values were derived from Cochran's Q test for heterogeneity.

Risk of bias assessment and certainty of evidence evaluation are provided in the Tables S4 and S5.

4. Conclusion

This study quantified the cumulative absolute benefit of SGLT2 inhibition for cardiorenal outcomes as event‐free time, providing complementary information to conventional HR‐based analyses. Patients receiving SGLT2 inhibitors gained an average of 17 event‐free days by 30 months and 40 days by 42 months, although the latter estimate should be interpreted cautiously because it was driven solely by the extended follow‐up of CREDENCE.

RMST analysis does not define a minimal clinically important difference and cannot determine the biological onset of treatment effect. Nevertheless, whereas HRs primarily quantify the relative reduction in risk, RMST‐derived event‐free time provides an absolute and clinically intuitive measure of treatment benefit.

The included trials differed in their definitions of the composite renal outcome, which may have resulted in differential sensitivity to the acute eGFR dip and subsequent chronic eGFR decline associated with SGLT2 inhibition. However, between‐trial heterogeneity in RMST differences remained low to moderate across landmark time points.

This study has several limitations. First, this was a post hoc analysis that was not prospectively registered, and the landmark time points presented were selected for descriptive purposes rather than prespecified. Second, reconstructed IPD are inherently subject to reconstruction error and do not permit subgroup analyses, such as stratification by baseline eGFR. Third, the reconstruction algorithm assumes non‐informative censoring; if censoring is associated with prognosis, reconstruction accuracy may be affected. Fourth, RMST estimates depend on the selected truncation time, and the magnitude of event‐free time may vary with follow‐up duration. Finally, the study population was limited predominantly to CKD patients with albuminuria, and caution is warranted when generalizing these findings to broader CKD populations.

Author Contributions

Conceptualization: Shotaro Yoshida. Methodology: Shotaro Yoshida, Takahiro Nakamura and Makiko Oda. Data curation and analysis: Shotaro Yoshida and Mashiro Iwatsubo. Writing – original draft: Shotaro Yoshida. Writing – review and editing: Shotaro Yoshioda, Mashiro Iwatsubo, Takahiro Nakamura, Makiko Oda and Yusuke Sakamaki. Supervision: Yusuke Sakamaki.

Funding

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: The complete search strategies.

Table S2: The maximum absolute difference between reconstructed and digitized Kaplan–Meier survival probabilities.

Table S3: Comparison of reconstructed and published HR (95% CI).

Table S4: Risk of bias assessment of included RCTs (Cochrane RoB 2).

Table S5: GRADE evidence profile.

Figure S1: PRISMA flow diagram.

Figure S2: Reconstructed Kaplan–Meier curves of the CREDENCE, DAPA‐CKD, EMPA‐KIDNEY trials.

Figure S3: Pooled RMST difference curve for cardiorenal composite endpoint extended to full follow‐up data of CREDENCE trial (42 months).

Figure S4: Pooled RMST difference curve for CV death (CREDENCE, EMPA‐KIDNEY).

Yoshida S., Iwatsubo M., Nakamura T., Oda M., and Sakamaki Y., “Time‐Dependent Accrual of Cardiorenal Benefit With SGLT2 Inhibitors in Chronic Kidney Disease: A Reconstructed Individual Patient Data Meta‐Analysis of Randomized Trials,” Diabetes, Obesity and Metabolism 28, no. 11 (2026): 10857–10861, 10.1111/dom.71195.

Data Availability Statement

Data derived from public domain resources.