Higher diversity of anti-diabetic therapy is associated with lower risk of major adverse kidney events in diabetes

Abstract

BACKGROUND

Diverse anti-diabetic combinations are widely indicated for uncontrolled hyperglycemia to avoid microvascular complications including kidney injury. Current studies on combination therapy’s long-term renal risk yield conflicting results, largely limited by inadequate confounder adjustment. We hypothesized combination anti-diabetic therapy reduces major adverse kidney events (MAKE) vs monotherapy in diabetic patients.

AIM

To investigate the association between the diversity of anti-diabetic therapy and long-term renal outcomes in patients with diabetes.

METHODS

We included participants with diabetes and glycated hemoglobin A1c ≥ 7% from the UK Biobank (UKB) and Chongqing Diabetes Registry (CDR). Using a target trial emulation design, we compared the effect of monotherapy vs combination therapy on MAKE (composite of end-stage renal disease and kidney disease death). Intention-to-treat and per-protocol analyses were conducted in UKB and CDR, respectively.

RESULTS

Propensity score matching identified 3475 monotherapy-combination therapy pairs in the UKB and 261-808 pairs in the CDR. Over median 13.4 years in UKB, combination therapy was associated with lower MAKE risk vs monotherapy (hazard ratio 0.65; 95% confidence interval: 0.46-0.92). In CDR, when compared with patients who maintained monotherapy, patients who maintained combination therapy had significantly lower risk of MAKE (hazard ratio 0.24; 95% confidence interval: 0.11-0.55) although their glycated hemoglobin A1c level was similar at baseline and during follow up. Sensitivity analyses excluding renoprotective medications showed consistent results.

CONCLUSION

Beyond glycemic control, a greater diversity of anti-diabetic therapy is associated with lower risk of MAKE in diabetic patients.

Key Words: Anti-diabetic therapy; Combination therapy; Major adverse kidney events; Target trial emulation; Type 2 diabetes mellitus

Core Tip: Greater diversity of anti-diabetic therapy improves glycemic control, while its impact on kidney outcomes remains controversial. In two propensity-matched cohorts, combination therapy was associated with a 35%-76% lower major adverse kidney events risk vs monotherapy, independent of glycemic control and renoprotective medications. These findings support a potential association between greater diversity of anti-diabetic therapy and lower major adverse kidney events risk in patients with diabetes.

INTRODUCTION

A higher diversity of anti-diabetic therapy is often prescribed for diabetic patients with uncontrolled hyperglycemia[1]. Guidelines recommend initiating combination therapy when monotherapy fails to maintain glycated hemoglobin A1c (HbA1c) below 7%[2-4]. As the major objective of glycemic control is to prevent diabetic microvascular complications[5], whether the increasing diversity of anti-diabetic therapy benefits long-term renal outcomes remains unclear.

The existing evidence regarding above relationship is inconsistent. Some studies found that combination therapies were associated with lower risk of chronic kidney disease (CKD) among patients with diabetes[6,7]. However, other studies showed that diabetic patients who received combination therapies had similar or even higher risk of CKD progression when compared to those who received monotherapy[8-10]. Those conflicting evidence may be ascribed, at least partially, to insufficient confounders adjustment, divergent criteria for assessing renal outcomes, and lack of validation in different cohorts.

The emulation design of a target randomized trial is a powerful tool to control confounders[11]. Using the target trial emulation framework in two large-scale cohorts [the UK Biobank (UKB) and the Chongqing Diabetes Registry (CDR)], we evaluated the association between combination anti-diabetic therapy and major adverse kidney events (MAKE), a widely accepted composite endpoint of end-stage renal disease and death from kidney disease. To further evaluate effects of glycemic control on the association between anti-diabetic therapy and MAKE, we also analyzed dynamic change of HbA1c across different exposures.

MATERIALS AND METHODS

Overall study design

We followed the target trial framework by Hernán and Robins[11] to emulate a randomized controlled trial comparing the effectiveness of monotherapy vs combination anti-diabetic therapy on the risk of MAKE, defined as a composite endpoint of end-stage renal disease and death from kidney disease. This target trial was emulated separately in two cohorts: The UKB and the CDR. A schematic of the study design is presented in Figure 1. We emulated the target trial in both cohorts by following the specified protocols to design the cohorts, define exposure definitions and outcomes, select confounders, and specify the analytic approaches. The protocol for the target trial in the UKB cohort, as well as key components of the emulated target trial in this cohort, are specified in Supplementary Table 1; the corresponding protocol and key components for the CDR cohort are specified in Supplementary Table 2.

Figure 1 Study flowchart. Study flowcharts for the UK Biobank cohort (left) and the Chongqing diabetes registry (right). HbA1c: Glycated hemoglobin A1c; PSM: Propensity score matching; MAKE: Major adverse kidney event; ESRD: Endstage renal disease.

Eligibility criteria

This study used data from two independent cohorts: The UKB and the CDR. UKB recruited more than 500000 participants aged 37 years to 73 years from 22 assessment centers across England, Scotland, and Wales between 2006 and 2010. Details of the study design and data collection have been described previously[12,13]. The current study was approved by UKB with an ID of 103654. Among the 502241 participants, a total of 27573 participants with baseline comorbid diabetes were initially screened.

Inclusion criteria: (1) Age ≥ 18 years; (2) Diagnosis of diabetes; (3) Treatment with at least one anti-diabetic agent at baseline; and (4) Baseline HbA1c ≥ 7%.

Exclusion criteria: MAKE at baseline, missing baseline anti-diabetic medication data or loss to follow-up. A total of 9370 eligible participants were included.

The CDR registered at ClinicalTrials.gov (ID: NCT03692884), was established in December 2017 and enrolled 3841 participants through March 2024, demonstrating a median follow-up duration of 4.0 years and a low loss-to-follow-up rate of 0.5% as of the latest update[14,15]. Data were collected via standardized questionnaires, electronic medical records, and annual follow-ups. The collected data covered demographic characteristics, medical history, lifestyle factors, anthropomorphic measurements, and laboratory test results, including HbA1c, renal function, and lipid profile. The study was approved by the Ethics Committee of the First Affiliated Hospital of Chongqing Medical University and that it conforms to the provisions of the Declaration of Helsinki (as revised in Fortaleza, Brazil, October 2013) (approval No. 2018-042), and all participants provided written informed consent. The inclusion criteria were consistent with those of the UKB. A total of 3841 patients were initially screened, and 2751 eligible participants were enrolled.

Definition of exposure

Monotherapy was defined as the use of only one anti-diabetic medication, while combination therapy was defined as the use of two or more anti-diabetic medications. In the UKB cohort, no follow-up data were available for medication information; only baseline anti-diabetic medication data were recorded. Among the eligible participants, 4964 received monotherapy and 4406 received combination anti-diabetic therapy at baseline. Given that its recruitment period was from 2006 to 2010, no patients had received dipeptidyl peptidase-4 (DPP4) inhibitors, glucagon-like peptide-1 (GLP-1) receptor agonists, or sodium-glucose cotransporter 2 (SGLT2) inhibitors. Detailed information on the use of each class of anti-diabetic medications is provided in Supplementary Table 3.

In the CDR, participants’ anti-diabetic medication data were recorded in detail both at baseline and during follow-up. Monotherapy was defined as initiating one anti-diabetic medication at enrollment and maintaining one medication at either an endpoint event or the last follow-up, while combination therapy was defined as initiating 2 or ≥ 3 medications at enrollment and maintaining the same number at the aforementioned time points. A total of 322 participants maintained monotherapy, and 1046 participants maintained combination therapy throughout the study. Information on their medication use is presented in Supplementary Table 4.

In CDR, prescribed daily doses were systematically documented besides medication classes. Defined Daily Doses (DDDs) were calculated following the World Health Organization Anatomical Therapeutic Chemical/DDD guidelines (2024 version)[16]. Medications classified under Anatomical Therapeutic Chemical code A10 (antidiabetic drugs) were categorized into subclasses, including A10A (insulins and analogs) and A10BJ (GLP-1 receptor agonists). For long-acting GLP-1 receptor agonists (e.g., dulaglutide 1.5 mg/week), the weekly dose was divided by 7 days to calculate the equivalent daily dose (e.g., 1.5 mg per week/7 days = 0.21 mg per day). Total DDDs per participant were calculated as the arithmetic sum of DDDs for all anti-diabetic medications, irrespective of drug class. For example, participants using metformin (daily dose 2000 mg; DDD 2000 mg, DDDs = 1.0) combined with sitagliptin (daily dose 100 mg; DDD 100 mg, DDDs = 1.0) had a total DDDs of 2.0 (Supplementary Table 5).

Determination of outcome events

The primary outcome was MAKE, defined as meeting at least one of the following criteria[17]: (1) Progression to end-stage renal disease, defined as an estimated glomerular filtration rate (eGFR) ≤ 15 mL/minute/1.73 m² (calculated using the CKD Epidemiology Collaboration Equation); initiation of dialysis (hemodialysis or peritoneal dialysis); or undergoing kidney transplantation; and (2) Death attributed to kidney disease, with cause of death adjudicated by an independent endpoint committee and coded according to the International Classification of Diseases, Tenth Revision codes N17-N19.

Covariates

The covariates were selected based on clinical knowledge and informed by previous randomized controlled trials. The covariates included demographic characteristics (sex, age and ethnicity), lifestyle factors (smoking status and alcohol consumption), comorbidities (hypertension, cardiovascular disease, cerebrovascular disease), clinical variables (diabetes duration, HbA1c, cholesterol, triglyceride, low density lipoprotein cholesterol, high-density lipoprotein cholesterol, uric acid, and eGFR), drug history (angiotensin-converting enzyme inhibitor, angiotensin-receptor blocker, mineralocorticoid receptor antagonist, and diuretics). For covariates with missing data (proportion < 5%), multiple imputation was performed with five iterations using the predictive mean matching algorithm in the R mice package (version 3.16.0).

Target trial emulation

The protocols for the emulated randomized controlled trial are detailed in Supplementary Tables 1 and 2. Treatment assignment was emulated based on the baseline anti-diabetic regimen in both cohorts: Participants were assigned to the monotherapy group or the combination therapy group upon fulfilling all eligibility criteria. Time zero was defined as the date when eligibility was met and the assigned treatment was initiated, corresponding to the baseline assessment date in the UKB cohort and the enrollment visit date in the CDR cohort. The end of follow-up was defined as the time from baseline assessment (2006-2010) until the occurrence of MAKE or the latest data extraction date (July 2022) in the UKB, and from baseline until MAKE occurrence or the end of follow-up (March 2024) in the CDR. To estimate causal contrasts of interest, we used the Kaplan-Meier estimator to obtain cumulative incidences for each treatment strategy within 5 years of follow-up, and Cox proportional hazards analyses to compare time-to-event rates on a daily basis over the follow-up period since baseline.

In the UKB, patients who received monotherapy or combination therapy at baseline were matched at a 1:1 ratio using propensity score matching to balance baseline covariates. The propensity score was estimated using logistic regression, and pairs were formed using greedy nearest-neighbor matching with a caliper width of 0.05[18]. The intention-to-treat effect on the risk of MAKE was estimated in the emulated target trials within the UKB.

In the CDR, we adopted the 1:5 nearest neighbor propensity score matching with replacement, with a propensity score estimated using logistic regression of the treatment on the covariates. The caliper value is set at 0.05 times the SD of the propensity score. The per protocol effect on the risk of MAKE was estimated in the emulated target trials within the CDR. For balance diagnostics, standardized mean difference < 0.05 was defined as the threshold for acceptable covariate balance after matching.

Pre-matching baseline characteristics and standardized mean difference values are provided in Supplementary Tables 6 and 7, and post-matching balance results are presented in Table 1.

Table 1 Baseline characteristics of participants in UK Biobank and Chongqing Diabetes Registry cohorts after propensity score matching, n (%).

Characteristic	UK biobank			Chongqing diabetes registry
	Monotherapy (n = 3475)	Combination therapy (n = 3475)	SMD	Monotherapy (n = 261)	Combination therapy (n = 808)	SMD
Age, year	61.00 (55.00, 65.00)	61.00 (55.00, 65.00)	-0.014	59.78 (48.56, 67.42)	58.41 (52.21, 66.79)	0.028
Female sex	1270 (36.5)	1311 (37.7)	0.025	107 (41.0)	322 (39.9)	0.034
Ethnicity (White)	2991 (86.1)	2983 (85.8)	-0.006	-	-	-
Diabetes duration, year	8.89 (5.05, 16.76)	9.00 (5.16, 14.16)	0.018	8.86 (2.13, 15.12)	9.95 (4.05, 15.02)	0.000
Current smoker	1591 (45.8)	1611 (46.4)	-0.012	107 (41.0)	340 (42.1)	-0.007
Current drinker	1641 (47.2)	1619 (46.6)	-0.013	61 (23.4)	186 (23.0)	-0.006
Hypertension	2317 (66.7)	2303 (66.3)	-0.009	93 (35.6)	343 (42.5)	-0.019
CVD	691 (19.9)	663 (19.1)	-0.020	6 (2.3)	24 (3.0)	0.004
CeVD	294 (8.5)	290 (8.3)	-0.004	9 (3.4)	25 (3.1)	0.017
BMI, kg/m2	31.00 (27.67, 35.18)	31.31 (28.10, 35.33)	0.048	23.51 (20.81, 25.50)	24.03 (22.13, 26.12)	-0.006
HbA1c %	7.85 (7.36, 8.65)	7.89 (7.39, 8.70)	0.009	9.30 (7.80, 11.30)	9.25 (8.00, 10.80)	0.044
Cholesterol, mmol/L	4.28 (3.76, 4.93)	4.25 (3.73, 4.91)	-0.019	4.23 (3.47, 5.10)	4.27 (3.62, 5.00)	-0.012
Triglycerides, mmol/L	1.93 (1.29, 2.78)	1.96 (1.35, 2.80)	0.000	1.34 (0.96, 2.04)	1.48 (1.02, 2.28)	-0.029
HDL cholesterol, mmol/L	1.11 (0.94, 1.31)	1.10 (0.95, 1.30)	-0.015	1.13 (0.93, 1.39)	1.08 (0.90, 1.29)	-0.011
LDL cholesterol, mmol/L	2.54 (2.15, 3.02)	2.53 (2.13, 3.01)	-0.017	2.53 (1.89, 3.20)	2.56 (2.01, 3.26)	-0.012
Uric acid, μmol/L	305.20 (254.60, 369.05)	309.00 (258.95, 363.95)	-0.006	304.00 (245.00, 381.00)	311.00 (251.00, 385.25)	-0.034
eGFR, mL/minute/1.73 m2	93.76 (81.60, 100.92)	93.66 (79.82, 101.03)	0.018	85.38 (64.66, 101.54)	87.07 (68.50, 100.30)	-0.003
RAS inhibitors	680 (19.6)	685 (19.7)	0.004	47 (18.0)	206 (25.5)	-0.028
Diuretic	720 (20.7)	733 (21.1)	-0.022	10 (3.8)	45 (5.6)	-0.048
MRA	56 (1.6)	47 (1.4)	-0.004	1 (0.4)	2 (0.2)	-0.014

Continuous variables are presented as median (25^th and 75^th percentiles) and categorical variables as n (%). SMD: Standardized mean difference; CVD: Cardiovascular disease; CeVD: Cerebrovascular disease; BMI: Body mass index; HbA1c: Glycated hemoglobin A1c; HDL: High-density lipoprotein; LDL: Low-density lipoprotein; eGFR: Estimated glomerular filtration rate; RAS: Renin-angiotensin system; MRA: Mineralocorticoid receptor antagonist.

Statistical analysis

Normality of continuous variables was assessed using the Kolmogorov-Smirnov test. Variables following a normal distribution were presented as mean SD, non-normally distributed variables as median (25^th percentile, 75^th percentile), and categorical variables as n (%).

Cox proportional hazards regression models were used to evaluate associations of different treatment regimens with MAKE risk across both cohorts, with follow-up time in years. Person-time was calculated from baseline assessment (2006-2010) until the occurrence of MAKE or the latest data extraction date (July 2022) in UKB; in CDR, it was calculated from baseline until MAKE occurrence or the end of follow-up (March 2024). The proportional hazards assumption was not violated (P > 0.05). To visually compare MAKE risks across treatment regimens, we generated Kaplan-Meier curves for monotherapy and combination therapy in both cohort.

We conducted several sensitivity analyses. In UKB (where data on SGLT2 inhibitors and GLP-1 receptor agonists use were unavailable), sensitivity analyses were restricted to the adjustment for or exclusion of baseline users of angiotensin-converting enzyme inhibitors/angiotensin receptor blockers and mineralocorticoid receptor antagonists. In the CDR, we sequentially adjusted for angiotensin-converting enzyme inhibitors/angiotensin II receptor blockers, mineralocorticoid receptor antagonists, SGLT2 inhibitors, and GLP-1 receptor agonists at baseline and during follow-up. Participants who used the aforementioned medications at baseline and during follow-up were also sequentially excluded from the analyses. In addition, we performed a competing-risk analysis, treating death from non-kidney causes as a competing event.

All analyses were conducted in R (version 4.3.1), with statistical significance defined as a two-tailed P value < 0.05. The statistical methods of this study were reviewed by Bin Peng from Chongqing Medical University.

RESULTS

After propensity score matching in the UKB, 3475 participants who received monotherapy at baseline were matched to 3475 participants who received combination therapy. In the CDR, 261 participants who maintained continuous monotherapy were matched to 808 participants who maintained continuous combination therapy. Baseline characteristics of participants in the monotherapy and combination therapy groups, before and after matching, in both the UKB and CDR cohorts are presented in Table 1, Supplementary Tables 6 and 7. All baseline characteristics were well balanced after matching with standardized mean difference less than 0.05.

In the UKB, MAKE occurred in 83 patients (2.4%) in the monotherapy group and 54 patients (1.6%) in the combination therapy group during a median follow-up of 13.4 years. The participants receiving combination anti-diabetic therapy had a 35% lower risk of MAKE than those receiving monotherapy [hazard ratio (HR) = 0.65, 95% confidence interval (CI): 0.46-0.92] (Table 2, Figure 2A). This relationship remained similar when adjusting for or excluding baseline users of angiotensin-converting enzyme inhibitors/angiotensin II receptor blockers or mineralocorticoid receptor antagonists, with HRs ranging from 0.59 to 0.65 in all models (all P < 0.05) (Table 2). In a competing-risk analysis accounting for death from other causes, the association was consistent (Supplementary Table 8).

Figure 2 Risk of major adverse kidney events, changes of hemoglobin A1c, and hypoglycemic events among patients who received monotherapy and combination anti-diabetic therapy. ^aP < 0.05. A: Kaplan-Meier curves of major adverse kidney events (MAKE) in UK biobank using a propensity score-matched method. Monotherapy: With a single anti-diabetic medication. Combination therapy: With ≥ 2 anti-diabetic medications. The MAKE is a composite of end-stage renal disease (defined as estimated glomerular filtration rate ≤ 15 mL/minute/1.73 m², initiation of dialysis, or undergoing kidney transplantation) or death attributed to kidney disease; B: Kaplan-Meier curves of MAKE in Chongqing Diabetes Registry using an emulation design of randomized target trial; C: Follow-up hemoglobin A1c levels between monotherapy and combination therapy groups in Chongqing Diabetes Registry; D: Self-reported hypoglycemic events between monotherapy and combination therapy groups in Chongqing Diabetes Registry. Hypoglycemic events include self-monitored fasting blood glucose levels < 3.9 mmol/L, or the presence of hypoglycemic symptoms (e.g., palpitations, sweating) without blood glucose measurement, which resolve after food intake.

Table 2 Effects of combination anti-diabetic therapy vs monotherapy on risk of major adverse kidney events.

Characteristic	Case/total		HR (95%CI)
	Monotherapy	Combination therapy
UK Biobank: Intention-to-treat analysis
Unadjusted model	83/3475	54/3475	0.65 (0.46-0.92)
Adjusting for ACEI/ARB and MRA use at baseline	83/3475	54/3475	0.65 (0.46-0.92)
Excluding ACEI/ARB users at baseline	68/2795	40/2790	0.65 (0.46-0.93)
Excluding MRA users at baseline	78/3419	51/3428	0.59 (0.40-0.87)
Chongqing Diabetes Registry: Per-protocol analysis
Unadjusted model	19/261	13/808	0.24 (0.11-0.55)
Adjusting for ACEI/ARB, MRA, SGLT2i, and GLP-1RA use at baseline or during follow-up	19/261	13/808	0.37 (0.16-0.88)
Excluding ACEI/ARB users at baseline or during follow-up	9/157	4/373	0.25 (0.07-0.95)
Excluding MRA users at baseline or during follow-up	19/258	12/797	0.24 (0.11-0.55)
Excluding SGLT2i users at baseline or during follow-up	19/203	6/288	0.30 (0.10-0.87)
Excluding GLP-1RA users at baseline or during follow-up	19/251	9/640	0.22 (0.09-0.56)

The major adverse kidney events are a composite of end-stage renal disease (defined as estimated glomerular filtration rate ≤ 15 mL/minute/1.73 m², initiation of dialysis, or undergoing kidney transplantation) or death attributed to kidney disease. Monotherapy: With a single anti-diabetic medication. Combination therapy: With ≥ 2 anti-diabetic medications. HR: Hazard ratio; CI: Confidence interval; ACEI: Angiotensin-converting enzyme inhibitor; ARB: Angiotensin receptor blocker; MRA: Mineralocorticoid receptor antagonist; SGLT2i: Sodium-glucose cotransporter-2 inhibitor; GLP-1RA: Glucagon-like peptide-1 receptor agonist.

In the CDR, the prevalence of MAKE was higher among participants maintained monotherapy compared with those maintained combination therapy (7.3% vs 1.6%, P < 0.001). Participants maintained combination therapy had a 76% lower risk of MAKE than those maintained monotherapy (HR = 0.24, 95%CI: 0.11-0.55) (Table 2, Figure 2B). Similar risk reductions were observed after additional adjustment for use of angiotensin-converting enzyme inhibitors/angiotensin II receptor blockers, mineralocorticoid receptor antagonists, SGLT2 inhibitors, and GLP-1 receptor agonists at baseline and throughout follow-up (HR = 0.37, 95%CI: 0.16-0.88), or even after excluding the users of SGLT2 inhibitors or GLP-1 receptor agonists (Table 2). In a competing-risk analysis accounting for death from other causes, the association was consistent (Supplementary Table 8).

We observed baseline and follow-up HbA1c levels at 2^nd and 4^th year were similar in participants maintained combination therapy compared to those maintained monotherapy (all P > 0.05) (Figure 2C, Supplementary Table 9). A significantly lower incidence of self-reported hypoglycemic events was observed in participants maintained on combination therapy than in those on monotherapy at both the 2nd and 4th years of follow-up (2^nd year: P = 0.004; 4^th year: P = 0.018) (Figure 2D, Supplementary Table 9). These findings may suggest the reduced MAKE risk associated with combination therapy may be independent of glucose lowering.

Compared to the monotherapy group, the risk of MAKE was significantly lower among patients who maintained 2 types of anti-diabetic medications (dual therapy) and maintained ≥ 3 types of anti-diabetic medications (polytherapy) (Supplementary Figure 1A). The 4-knot restricted cubic spline model indicated a linear inverse association between anti-diabetic medication DDDs and the risk of MAKE (Supplementary Figure 1B), and one-unit increment of DDDs was associated with a 58% lower risk of MAKE (HR = 0.42, 95%CI: 0.22-0.80).

DISCUSSION

Using an emulation design of randomized target trial to control potential confounders, we firstly demonstrated that the combination anti-diabetic medications was associated with a lower risk of MAKE when compared to monotherapy, and the main results remained similar even after excluding the effects of SGLT2 inhibitors and GLP-1 receptor agonists. These findings highlight that beyond the glycemic control, the guideline recommended strategy of combination anti-diabetic therapy is helpful for delaying diabetic kidney disease (DKD) progression.

The existing evidence regarding the association between diversity of anti-diabetic therapy with long-term renal outcomes remains controversial. On one hand, a study included 16248 patients with a follow-up of 3.2 years, and the results showed that when compared with the monotherapy of SGLT2 inhibitor, adding metformin on SGLT2 inhibitor significantly reduced the risk of composite renal outcomes[6]. Another retrospective cohort study involving 1512 patients with DKD demonstrated that DPP4 inhibitors added to metformin and/or α-glucosidase inhibitors significantly prevent renal function decline[7]. On the other hand, a prespecified analysis of the FLOW trial involving 550 participants with DKD observed no significant improvement in the composite renal outcome with the addition of semaglutide among patients who were already receiving SGLT2 inhibitors[8]. A retrospective cohort study of 1715 patients with diabetes found that combination regimens consisting of metformin with sulfonylureas or metformin with insulin were not associated with a significant reduction in the risk of elevated serum creatinine compared with sulfonylurea monotherapy, even after adjustment for confounders such as baseline glucose levels and albuminuria[9]. In addition, a study involving 624 patients with DKD showed that combination therapy with insulin and an SGLT2 inhibitor was associated with a higher risk of albuminuria stage progression or eGFR decline compared to SGLT2 inhibitor monotherapy[10]. In the current study, we used a randomized target trial emulation design to mitigate potential confounding, applied MAKE as a hard endpoint, and cross-validated in two ethnically distinct cohorts. Our results confirmed that combination therapy was associated with a significantly reduced risk of MAKE than monotherapy.

Whilst guidelines recommend combination anti-diabetic therapy to achieve glycemic control, the evidence for its effects on renal outcomes seems not to be strong enough[2-4]. In our study, similar HbA1c levels were documented in participants with combination therapy compared to those receiving monotherapy across baseline, 2-year follow-up and 4-year follow-up. We further observed a lower hypoglycemia risk associated with combination therapy. These findings indicate that the decreased MAKE risk with combination therapy may be independent of glycemic control. Previously, a study demonstrated a reduction in cardiovascular risk with quadruple therapy compared to dual therapy, even when achieving comparable glycemic control that measured by follow-up HbA1c levels[19]. These evidences suggest that several mechanisms independent of glycemic control may contribute to the reduction in MAKE risk associated with increasing diversity of anti-diabetic therapy, and one of potential interpretations might be the minimizing side effect of hypoglycemia on renal decline among patients with diabetes[20].

The protective effect of anti-diabetic medication diversity on MAKE is also likely attributable to the synergy between distinct drug mechanisms. In our analysis of Supplementary Tables 3 and 4, the use of biguanide was more frequent among patients receiving combination therapy than monotherapy: 95.9% vs 51.4% in UKB, and 87.9% vs 21.1% in CDR. In UKB, the proportion of using thiazolidinediones was 1.2% and 19.6% among participants receiving monotherapy and combination therapy, respectively. Considering that insulin resistance may accelerate glomerular hyperfiltration and tubulointerstitial inflammation[21-25], we speculated that the reduction in MAKE risk associated with increasing diversity of anti-diabetic therapy may rely on the improvement of insulin resistance by adding insulin enhancers of biguanide or thiazolidinediones. On the other hand, in CDR, the proportions of using SGLT2 inhibitors, DPP4 inhibitors, GLP-1 receptor agonists were 26.0%, 19.9%, and 10.5% for patients with combination therapy, respectively; while the proportions were all ≤ 1% among patients with monotherapy. Considering that an impaired GLP-1 signaling[26,27] and an enhanced SGLT2 activity are also involved in the progression of DKD[28,29], combinations of SGLT2 inhibitors, DPP4 inhibitors and GLP-1 receptor agonists may protect diabetic kidney via the synergistic effects of different mechanisms of action. Of note, our sensitivity analyses showed that even after excluding the uses of SGLT2i and GLP-1 receptor agonists that were commonly recognized as renal protective agents, the risk of MAKE also remained lower among patients with combination anti-diabetic medications when compared to monotherapy. These findings collectively indicate that the observed inverse correlation with kidney adverse events is a universal property of mechanistic interplay, rather than a consequence of specific anti-diabetic drug classes.

The strengths of our study lie in the emulation design of randomized target trial, large sample size and cross-ethnic validation between these two populations. Our analysis has several limitations. First, the UKB cohort lacks information on medication use during follow-up, restricting our analysis to the intention-to-treat effect. As a result, the impact of dynamic changes in medication use remains unexplored, and exposure misclassification may be introduced. However, the CDR cohort was designed to incorporate considerations of sustained medication use and employed a per-protocol analysis for outcome assessment, thereby addressing this shortcoming of the UKB. On the other hand, despite adjustment for multiple covariates, residual confounding remains a possibility from unmeasured factors such as incomplete diabetes severity characterization, medication adherence, and physician prescribing patterns. In addition, because our study is an observational emulation of a target trial, the observed associations between combination therapy and MAKE risk do not imply causation and warrant cautious interpretation. Future randomized controlled trials are warranted to elucidate the causal relationship between diverse antidiabetic therapies and MAKE.

CONCLUSION

Our study demonstrates that a greater diversity of anti-diabetic therapy is inversely associated with the risk of MAKE in patients with diabetes, independent of glycemic control and the use of established renoprotective medications. These findings further emphasize the importance of combination therapy and aggressive pharmacotherapeutic strategies in diabetes management for improving long-term renal outcomes. Of note, given the observational design, these associations cannot be interpreted as causal. Randomized controlled trials are required to verify whether combination therapy improves renal prognosis.

ACKNOWLEDGEMENTS

We thank the participants and staff of the UK Biobank and Chongqing Diabetes Registry. We also thank Bin Peng (Department of Statistics, School of Public Health, Chongqing Medical University) for the statistical review of this study.