Effect of sodium-glucose cotransporter-2 inhibitors on kidney outcomes of individuals with type 2 diabetes according to blood pressure levels

Abstract

Aims

Sodium–glucose cotransporter-2 (SGLT2) inhibitors have proved kidney-protective. Given that the SGLT2 inhibitors lower blood pressure (BP), the magnitude of their kidney benefits may vary depending on an individual’s BP. Therefore, we investigated whether baseline BP modifies the effect of SGLT2 inhibitors on kidney function.

Methods and results

This study included individuals with SGLT2 inhibitors or dipeptidyl peptidase-4 (DPP4) inhibitors newly prescribed for type 2 diabetes using a nationwide epidemiological cohort and performed propensity score matching (1:2). The primary outcome was the annual estimated glomerular filtration rate (eGFR) decline. We further investigated the interaction effect of systolic BP (SBP) at the time of prescription using a 3-knot restricted cubic spline model. We analysed 2148 individuals with SGLT2 inhibitor prescriptions and 4296 matched individuals with DPP4 inhibitor prescriptions. Overall, the annual eGFR decline was less pronounced in the SGLT2 inhibitor group than in the DPP4 inhibitor group (−1.32 mL/min/1.73 m² vs. −1.50 mL/min/1.73 m²). The treatment effect of SGLT2 inhibitors over DPP4 inhibitors was augmented with higher SBP (P for interaction = 0.0199). Further, after adjusting the definition of outcomes to a 30 or 40% reduction in eGFR, the advantages of SGLT2 inhibitors persisted, with a trend of augmented effect with higher SBP. Notably, annual eGFR decline was exacerbated for females presented with lower SBP when treated with SGLT2 inhibitors compared with DPP4 inhibitors.

Conclusion

This nationwide cohort analysis demonstrated that the kidney-protective effect of SGLT2 inhibitors could be modified by baseline SBP, highlighting the importance of patient selection by assessing their BP.

Graphical Abstract

Open in new tabDownload slide

Introduction

Randomized controlled studies have confirmed the renal protective benefits of sodium–glucose cotransporter-2 (SGLT2) inhibitors, underscoring their essential function in managing and preventing chronic kidney disease (CKD).^1–4 As a result, the use of SGLT2 inhibitors in medical settings is on the rise, and the determinants of their efficacy have emerged as an important area of clinical research.

The beneficial effects of SGLT2 inhibitors are attributable to multiple pathways. In addition to glucose-lowering effects, SGLT2 inhibitors improve arterial stiffness and vascular endothelial functions, which may be particularly relevant in patients with elevated baseline blood pressure (BP), contributing to their antihypertensive effects.^5–10 In individuals with diabetes, hypertension (or high BP) is common and can exacerbate glomerular hypertension, thereby impairing kidney function. Given the favourable vascular and metabolic actions of SGLT2 inhibitors, it is plausible that systemic BP status could influence their kidney-protective potential. However, it is not clear whether baseline BP modifies the effect of SGLT2 inhibitors on kidney function. We hypothesize that the kidney-protective effects of SGLT2 inhibitors may be more pronounced in cases with higher BP. Here, we used a nationwide health check-ups and claims dataset and investigated whether the kidney-protective effects of SGLT2 inhibitors, prescribed for individuals with type 2 diabetes alongside dipeptidyl peptidase-4 (DPP4) inhibitors, could be influenced by the BP at the initiation of the prescription.

Methods

Study design and data source

We conducted this investigation as a retrospective cohort study, employing the comprehensive DeSC database provided by DeSC Healthcare Inc. in Tokyo, Japan.^5–8 This database consolidates extensive health check-up records with administrative claims data spanning from April 2014 to November 2022. It encompasses data from three types of Japanese insurance: (i) association/union-administered health insurance for salaried employees working for relatively large companies in Japan; (ii) National Health Insurance for unemployed individuals aged <75 years; and (iii) the Advanced Elderly Medical Service System for older individuals aged ≥75 years. The DeSC database is noted for its extensive coverage and reliability, capturing a broad demographic spectrum in Japan, ranging from younger individuals to older individuals. It anonymizes and aggregates patient data from inpatient and outpatient services, facilitating longitudinal health tracking across multiple medical facilities. Diagnoses in the database are recorded using ICD-10 codes, and it includes regular health check-up data such as laboratory results, body measurements, and lifestyle questionnaires.

To mitigate biases associated with treatment selection and other uncontrolled variables, our analysis employed a new-user, active comparator study design using DPP4 inhibitors as a comparator, as illustrated in Supplementary material online, Figure S1. Dipeptidyl peptidase-4 inhibitors are prevalently used medications for diabetes in Japan.⁹ Dipeptidyl peptidase-4 inhibitors demonstrate HbA1c lowering effects comparable to SGLT2 inhibitors, whereas the effects of DPP4 inhibitors on cardiorenal outcomes are reported to be neutral.^9–12 Therefore, our study designated individuals with diabetes who were newly prescribed DPP4 inhibitors as the reference group, enabling a valid comparison of kidney outcomes attributable to SGLT2 inhibitors.

We extracted data for 24 795 individuals diagnosed with type 2 diabetes who had available estimated glomerular filtration rate (eGFR) data and were newly prescribed SGLT2 or DPP4 inhibitors at least 1 year after enrolment in their respective insurance coverages. We excluded people with a history of kidney replacement therapy (n = 57), those with eGFR below 15 mL/min/1.73 m² (n = 76), those with systolic BP (SBP) over 180 mmHg (n = 359), those with SBP under 90 mmHg (n = 43), and those missing data on smoking (n = 1887), alcohol consumption (n = 1340), urine protein (n = 61), or without a follow-up eGFR measurement (n = 9694). Ultimately, our study included 11 278 participants (as shown in Supplementary material online, Figure S2). In particular, patients with severely impaired kidney function (eGFR below 15 mL/min/1.73 m²) were excluded to maintain a consistent study population across the entire period because SGLT2 inhibitors were traditionally not indicated in such patients.

Ethics

This research was approved by the Ethics Committee at the University of Tokyo under the approval number 2021010NI and adhered to the ethical standards outlined in the Declaration of Helsinki. Given that all data utilized were anonymized, obtaining informed consent from participants was not required. Access to the DeSC database is granted to individuals who have procured it from DeSC Healthcare Inc.

Measurements and definitions

We examined health check-up data prior to prescribing SGLT2 or DPP4 inhibitors, which included measurements of body mass index (BMI), BP, HbA1c, LDL cholesterol, HDL cholesterol, triglycerides, and proteinuria levels (ranging from negative to 3+), determined through a urine dipstick. Data regarding smoking habits (current or non-current) and frequency of alcohol intake (daily or less) were also gathered via self-administered questionnaires during the health check-up. Additionally, from the claims records, we extracted information concerning diabetic complications [nephropathy (ICD-10 codes: E102, E112, E122, E132, E142), retinopathy (E103, E113, E123, E133, E143), and neuropathy (E103, E113, E123, E133, E143)] existing as of the date the SGLT2 or DPP4 inhibitors were prescribed. Medication details on the date these inhibitors were prescribed were also collected.

Outcomes

Data collection spanned from April 2014 to November 2022. The primary outcome of this study was defined as the rate of eGFR decline following the administration of SGLT2 or DPP4 inhibitors. This was estimated using a linear mixed-effects model, which utilized an unstructured covariance structure, as detailed in previous studies.^13,14

Propensity score matching

To evaluate the effects of initiating SGLT2 and DPP4 inhibitors for people with type 2 diabetes, we utilized a propensity score matching algorithm to establish a comparable cohort. The propensity scores for SGLT2 inhibitor users were calculated using a logistic regression model that considered various factors including age, gender, BMI, SBP, and diastolic BP, baseline eGFR, HbA1c, lipid levels (LDL cholesterol, HDL cholesterol, and triglycerides), lifestyle habits (smoking and alcohol consumption), diabetic complications (nephropathy, retinopathy, and neuropathy), medication use (including insulin, GLP-1 receptor agonists, biguanides, sulfonylureas, α-glucosidase inhibitors, thiazolidines, glinides, renin–angiotensin system inhibitors, β-blockers, calcium channel blockers, mineralocorticoid receptor antagonists, diuretics, and statins), and urine protein levels by dipstick. We conducted the matching at a 1:2 ratio to increase the precision, considering the high prevalence of DPP4i users in Japan,¹⁵ and employed a calliper width of 0.2 SD of the logit of the propensity score.¹⁶

Statistical analysis

Descriptive statistics were presented using the median (inter-quartile range) for continuous variables and count (percentage) for categorical variables. We employed a linear mixed-effects model with random intercepts and slopes, assuming an unstructured covariance structure, to assess the difference in eGFR changes between SGLT2 and DPP4 inhibitors. To investigate how SBP might alter these effects, we included terms for SBP, its interaction with the type of inhibitors (SGLT2 inhibitors or DPP4 inhibitors), and the individual annual eGFR slopes in our model. Systolic BP was modelled using a restricted cubic spline with three knots at 10, 50, and 90 percentiles, and we calculated the interaction P-value to explore whether SBP influenced the rate of eGFR decline differently in the two groups.

Additionally, we performed four sensitivity analyses: First, we executed subgroup analyses segmented by sex (female vs. male), age (≥50 vs. <50 years), eGFR at baseline (≥60 vs. <60 mL/min/1.73 m²), use of antihypertensive medications and renin–angiotensin system inhibitors, and urine protein status [negative (−/±) and positive (+1/+2/+3)]. Second, we assessed outcomes for individuals who maintained usage of SGLT2 inhibitors for more than 3 months. Third, we defined a 30 or 40% decrease in eGFR as an outcome, which has been widely employed in large-scale trials as validated endpoints for CKD progression.^17–19 We used Cox regression to compute hazard ratios (with 95% confidence intervals) for individuals treated with SGLT2 inhibitors, using those treated with DPP4 inhibitors as the reference group, and examined the impact of baseline SBP on these ratios. Fourth, we conducted the 1:1 propensity matching to account for the potential bias by using a one-to-many propensity matching strategy. Fifth, we limited the inclusion and observation periods before 2019 to exclude the confounding of acute kidney injury by COVID-19 infection.

A significance threshold was set at P < 0.05 for all tests. For the interaction, a criterion of P-value for interaction <0.10 was set to determine statistical significance. Statistical analyses were conducted using Stata v18 (StataCorp LLC, College Station, TX, USA).

Results

Clinical characteristics

Table 1 summarizes the clinical characteristics of study participants before and after propensity score matching. After 1:2 propensity score matching, 2148 well-balanced pairs were created. The median age was 65 (56–69) years for SGLT2 inhibitor users and 65 (55–69) years for DPP4 inhibitor users. In addition, 1403 (65.3%) individuals were men in SGLT2 inhibitor users, and 2837 (66.0%) individuals were men in DPP4 inhibitor users. The median eGFR was 71.8 (60.9–83.3) mL/min/1.73 m² in SGLT2 inhibitor users and 71.8 (61.1–83.1) mL/min/1.73 m² in DPP4 inhibitor users. The distribution of SBP after propensity score matching was summarized in Supplementary material online, Figure S3, and the median SBP was 132 (122–143) mmHg.

Change in estimated glomerular filtration rate among SGLT2 and dipeptidyl peptidase-4 inhibitors

The median follow-up period was 642 (389–991) days. Figure 1 shows the annual eGFR decline rate of individuals prescribed SGLT2 or DPP4 inhibitors according to the baseline SBP. Overall, the annual rate of eGFR decline was significantly reduced in the group prescribed SGLT2 inhibitors compared with DPP4 inhibitors (−1.32 mL/min/1.73 m² vs. −1.50 mL/min/1.73 m², P < 0.0001). In addition, the reduction of annual eGFR decline in the SGLT2 inhibitor group was augmented with higher SBP (P-value for interaction = 0.0199).

Sensitivity analysis

First, we conducted a subgroup analysis (Figure 2). The annual decline of eGFR was consistently reduced in the SGLT2 inhibitor group. Of note, a significant augmentation with higher SBP was observed in female individuals, people aged <50 years, those with eGFR ≥ 60 mL/min/1.73 m², and those not prescribed antihypertensive medications or renin–angiotensin system inhibitors at baseline.

Second, we included 10 339 individuals who continued to use the SGLT2 inhibitor or DPP4 inhibitor for more than 3 months. We found annual eGFR decline was attenuated in the SGLT2 inhibitor group, which was pronounced with higher SBP (P-value for interaction = 0.0692) (see Supplementary material online, Figure S4).

Third, we defined the kidney outcome as a decline in eGFR ≥ 30 and ≥40% decline. Overall, the decline in eGFR ≥ 30% and eGFR ≥ 40% was significantly lower in the SGLT2 inhibitor group than in the DPP4 inhibitor group (P = 0.0021 and P = 0.0072, respectively; Figure 3A and B). The hazard ratio of the eGFR decline in the SGLT2 inhibitor group varied according to BP levels, showing a significant reduction in individuals with SBP ranging from 130 to 150 mmHg (Figure 3C and D).

Fourth, we employed a 1:1 propensity matching strategy. After matching, 2148 pairs of DPP4 inhibitor and SGLT2 inhibitor users were included (see Supplementary material online, Table S1). The annual rate of eGFR decline was significantly reduced in the group prescribed SGLT2 inhibitors compared with DPP4 inhibitors (−1.32 mL/min/1.73 m² vs. −1.51 mL/min/1.73 m², P < 0.0001). The trend of the change in treatment effects of SGLT2 inhibitors across the baseline BP was consistent with the main analysis (P for interaction = 0.0685) (see Supplementary material online, Figure S5).

Fifth, we limited the inclusion and observation period before 2019 when the COVID-19 pandemic started. After 1:2 propensity matching, 1502 DPP4 inhibitor users and 751 SGLT2 inhibitor users were included (see Supplementary material online, Table S2). Consistent with the main result, the annual rate of eGFR decline was significantly reduced in the group prescribed SGLT2 inhibitors compared with DPP4 inhibitors (−1.13 mL/min/1.73 m² vs. −1.41 mL/min/1.73 m², P < 0.0001). The trend of the change in treatment effects of SGLT2 inhibitor across the baseline BP was comparable to the main analysis, without statistical significance (P for interaction = 0.111) (see Supplementary material online, Figure S6).

Discussion

Using a large epidemiological dataset, we analysed 11 278 individuals with type 2 diabetes who had newly taken SGLT2 inhibitors or DPP4 inhibitors, and substantiated the advantage of SGLT2 inhibitors on kidney outcomes over DPP4 inhibitors in our large cohort encompassing individuals having a diverse range of SBP. Further, the possible kidney-protective effects of SGLT2 inhibitors over DPP4 inhibitors were augmented in individuals with a higher SBP, particularly in individuals with preserved eGFR or non-use of antihypertensive medications at baseline. To the best of our knowledge, this is the first investigation to show the effect of SGLT2 inhibitors on kidney outcomes according to SBP using a large-scale real-world dataset (Graphical Abstract).

Several investigations are focusing on the impact of BP levels on the kidney benefits of SGLT2 inhibitors as sub-analyses of randomized controlled trials revealing the mixed results. Prespecified subgroup analyses of DAPA-CKD and EMPA-KIDNEY trials demonstrated no apparent interaction effect of BP on primary endpoints,^2,7 whereas the post hoc analysis of EMPA-REG OUTCOME suggested that BP might mediate these effects, secondary to the haematocrit.²⁰ These findings suggest that individual characteristics in these trials play a crucial role, emphasizing the need for further evaluations of the BP effect on kidney outcomes in diverse SGLT2 inhibitor users.

Our study introduces novelties and implications for clinical practice, distinguished from previous studies. This study is the first to examine the modification of kidney-protective effect of SGLT2 inhibitors by BP using large-scale real-world data that includes a wide range of patient profiles reflecting our clinical practice. As a result, SGLT2 inhibitors demonstrated a more favourable effect on the decline in eGFR compared with DPP4 inhibitors across a wide range of BP, and this possible advantage increased with higher BP. This result suggested that SGLT2 inhibitors could demonstrate greater kidney-related benefits for individuals with type 2 diabetes presenting with higher BP. SGLT2 inhibitors exert their kidney-protective effects through multiple mechanisms, such as the reduction in intraglomerular pressure through tubule-glomerular feedback, enhancement of tubular oxygenation, mitigation of inflammation, and increased stimulation of erythropoiesis.²¹ In addition, SGLT2 inhibitors are known to have the effect of lowering BP,^22–26 and we recently reported that SGLT2 inhibitors have also shown the potential to reduce the risk of developing hypertension.²⁷ It is reasonable to presume that this antihypertensive effect of SGLT2 inhibitors contributed to enhanced kidney protection for individuals with higher BP. Our results also showed that the advantage of SGLT2 inhibitors over DPP4 inhibitors was apparent across a broad spectrum of BP in terms of more clinically significant outcomes, such as a decrease in eGFR of 30 or 40%. A significant trend was observed in individuals with SBP ranging from 130 to 150 mmHg. However, due to limited observation periods and event numbers, these findings require further investigation.

Notably, in females or those not on antihypertensive medications, there was a trend towards accelerated eGFR decline in the group receiving SGLT2 inhibitors compared with those receiving DPP4 inhibitors when SBP was low. Females were older, less frequently used renin–angiotensin system inhibitors/beta blockers, and had lower urine protein levels compared with males (see Supplementary material online, Table S3). For these populations, the administration of SGLT2 inhibitors might have caused an excessive lowering of BP, potentially leading to deterioration in kidney function. Conversely, the kidney-protective effects of SGLT2 inhibitors were enhanced in hypertensive individuals who were young (age < 50 years) or not receiving antihypertensive medications and renin–angiotensin system inhibitors. Younger individuals are characterized by higher BMI and less frequent usage of antihypertensive medications (see Supplementary material online, Table S4). Clinicians would be encouraged to consider the use of SGLT2 inhibitors for these populations in hypertensive conditions.

Our results appear inconsistent with the sub-analyses of randomized controlled trials on patients with CKD, where no apparent modification of kidney-protective effect by BP was reported.^2,3 However, the present study also found the pronounced modification effect of BP in the individuals with baseline eGFR ≥ 60 mL/min/1.73 m², comprising 78% of the studied population. This suggests that the modification effect of BP could be significant for individuals with preserved kidney function, which might explain these discrepancies with the previous studies. Nevertheless, further investigation into the modification effect of BP on SGLT2 inhibitors is necessary.

We recognize several limitations in our investigation. First, the BP data were obtained from health check-up records, and therefore, it is possible that these measurements were not conducted using the most current standardized methods. Consequently, there is a chance that the BP values reported may not accurately represent actual BP levels. Second, given the observational cohort design of our study, we cannot eliminate the potential influence of unmeasured confounding variables. For example, we did not have data regarding the duration of diabetes, a factor that could significantly impact clinical outcomes. Third, we employed 1:2 propensity matching, considering that the prevalence of DPP4 inhibitors is higher than SGLT2 inhibitors in Japan. The strategy of one-to-many matching increases the precision of the study but sometimes could cause bias.¹⁵ We also conducted the 1:1 propensity matching as the sensitivity analysis, and the result was consistent with the main analysis. Fourth, our investigation period included the COVID-19 pandemic, and the acute kidney injury caused by COVID-19 infection could confound the results. We conducted the sensitivity analysis limiting the inclusion and observation periods before 2019, which showed consistent findings with the main analysis. Fifth, our database did not permit capturing the early, often transient eGFR dip seen shortly after SGLT2 inhibitor initiation. Future research with shorter-term lab measurements could help clarify whether the acute eGFR dip differs based on baseline BP. Sixth, the assessment of urinary albumin is important for predicting the benefits of SGLT2 inhibitors in kidney function.¹⁸ Unfortunately, our database lacked information on urinary albumin. In the subgroup analysis using urine dipstick tests, both urine protein positive and negative groups demonstrated a reduction in annual eGFR decline with SGLT2 inhibitors. Further study is needed to investigate the effect of urine albumin levels and baseline BPs on the kidney protection of SGLT2 inhibitors, ensuring a sufficient sample size. Finally, we could not perform a subgroup analysis based on the aetiological classification of nephropathy because our database lacks detailed clinical or pathological data. Future prospective cohorts or registries with more granular information will be critical to clarify these differences.

Perspective

Our large-scale epidemiological cohort analysis demonstrated that SGLT2 inhibitors had kidney-protective effects across a wide spectrum of BP, which was pronounced in individuals having a higher baseline SBP. On the other hand, the potential kidney-protective effect of SGLT2 inhibitors might be attenuated in people with lower BP. These results do not suggest an absence of benefit in individuals with lower BP; rather, they support a greater relative benefit among those with higher SBP. Clinicians should be encouraged to understand the potential heterogeneities and monitor kidney function after initiation of SGLT2 inhibitors. Further studies should confirm the modification effect of BP on SGLT2 inhibitors to explore the optimal usage of this epoch-making medication.

Supplementary material

Supplementary material is available at European Journal of Preventive Cardiology.

Author contribution

T.J. and H.K. were involved in the conception, design, and conduct of the analysis and interpretation of the results. Y.S. and A.O. contributed to the conduct of the analysis and interpretation of the results. T.A., T.K., K.F., N.T., H.M., K.H., T.Y., K.N., I.K., H.Y., M.N., and N.T. contributed to the interpretation of the results and critical revision of the manuscript. All gave final approval and agreed to be accountable for all the aspects of work, ensuring integrity and accuracy.

Funding

This work was supported by grants from the Ministry of Health Labour and Welfare Japan (23AA2003) and the Ministry of Education, Culture, Sports, Science and Technology, Japan (20H03907, 21H03159, and 21K08123). The funding sources had nothing regarding the current study.

Data availability

The DeSC database is available for purchase from DeSC Healthcare Inc.

References

Author notes