The effect of sodium‐glucose cotransporter 2 inhibitors on HbA1c variability and cardiovascular and renal adverse outcome in patients with T2DM

Abstract

Aims

To compare the effectiveness of sodium‐glucose cotransporter 2 (SGLT2) inhibitors and dipeptidyl peptidase 4 (DPP‐4) inhibitors in reducing haemoglobin A1c (HbA1c) variability and improving cardiovascular and renal outcomes in patients with type 2 diabetes mellitus (T2DM) and high HbA1c variability.

Methods

This territory‐wide cohort study involved patients with T2DM and an HbA1c variability score (HVS) >60% who initiated SGLT2 inhibitors or DPP‐4 inhibitors in Hong Kong between 2015 and 2022. Propensity score (PS) matching was used to adjust for confounders. The primary outcome was post‐treatment HVS within 3 years. Secondary outcomes included major adverse cardiovascular events (MACE) and serious renal events (SRE).

Results

Among 20,205 T2DM patients with a baseline HVS >60%, 4,612 SGLT2 inhibitor users were 1:1 matched with DPP‐4 inhibitor users. When referencing the 0%–20% quintile, patients initiating SGLT2 inhibitors versus DPP‐4 inhibitors exhibited a reduced likelihood of being in higher HVS quintiles [21%–40%: odds ratio (OR) 0.76, 95% confidence interval (CI) 0.66–0.88; 41%–60%: OR 0.57, 95% CI 0.50–0.65; 61%–80%: OR 0.49, 95% CI 0.42–0.56; and 81%–100%: OR 0.40, 95% CI 0.34–0.47]. SGLT2 inhibitors were associated with a reduced risk of MACE [hazard ratio (HR) 0.69; 95% CI 0.60–0.79] and SRE (HR 0.71; 95% CI 0.63–0.80) compared to DPP‐4 inhibitors.

Conclusion

In patients with high HbA1c variability, SGLT2 inhibitor initiation was associated with superior effectiveness in reducing HbA1c variability compared to DPP‐4 inhibitors. The initiation of SGLT2 inhibitors versus DPP‐4 inhibitors was linked to significantly reduced cardiovascular and renal adverse events.

1. INTRODUCTION

Long‐term glycaemic variability, usually measured by visit‐to‐visit haemoglobin A1c (HbA1c) variability, is linked to an increased risk of cardiovascular and renal complications and mortality in patients with type 2 diabetes mellitus (T2DM). ¹ , ² , ³ , ⁴ , ⁵ , ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ High HbA1c variability is associated with a two‐to‐three‐fold increased risk of major adverse cardiovascular events (MACE), progression of chronic kidney diseases, and all‐cause mortality, even in individuals with normal HbA1c levels. ³ Despite the recognized need for heightened medical attention for patients with T2DM and high HbA1c variability, there remains a lack of evidence regarding the optimal glycaemic controlling strategy for this population.

Sodium‐glucose cotransporter‐2 (SGLT2) inhibitors and dipeptidyl peptidase‐4 (DPP‐4) inhibitors are novel oral antidiabetic drugs that are commonly prescribed to patients with T2DM. ¹¹ , ¹² Although comparative studies have shown that SGLT2 inhibitors possess higher glucose‐lowering efficacy than DPP‐4 inhibitors, ¹² there remains a lack of conclusive evidence regarding the superior class for managing HbA1c variability. Multiple randomized controlled trials (RCTs) have acknowledged the cardiovascular and renal protective benefits of SGLT2 inhibitors. ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ However, individuals with high HbA1c variability are commonly excluded from RCTs due to not meeting the stringent inclusion criteria, thereby limiting the comprehensive understanding of the effects of SGLT2 inhibitors on this vulnerable patient subgroup.

In this territory‐wide cohort study, we aimed at: (1) comparing the effectiveness of SGLT2 inhibitors versus DPP‐4 inhibitors on reducing HbA1c variability; (2) evaluating the cardiovascular and renal outcomes associated with SGLT2 inhibitors versus DPP‐4 inhibitors in patients with T2DM and high HbA1c variability. We employed the HbA1c variability score (HVS) as the metric for quantifying HbA1c variability. HVS serves as a measure of the frequency of HbA1c oscillations and is more straightforward and translatable than the standard deviation of HbA1c (SD_HbA1c) and coefficient of variation of HbA1c (CV_HbA1c) in clinical settings. ⁴

2. MATERIALS AND METHODS

2.1. Study design and data source

We conducted a retrospective cohort study of de‐novo users of SGLT2 inhibitors versus DPP‐4 inhibitors based on the data from the Clinical Data Analysis Reporting System (CDARS), which is a territory‐wide database developed by the Hong Kong Hospital Authority (HA). The HA is the sole provider of public healthcare services in Hong Kong, which provides around 90% of inpatient services for a population of 7.3 million residents. ²⁰ Previously published epidemiological studies have validated the accuracy and authenticity of the data available in CDARS. ²¹ , ²² All the personal data (name and Hong Kong identification number) were deidentified in CDARS, and unique reference numbers were assigned to individuals. The study was approved by the institutional review board of the University of Hong Kong and the West Cluster of the Hong Kong Hospital Authority.

2.2. Study population

All patients with T2DM aged 18 or above and with HVS >60% who initiated an SGLT2 inhibitor or a DPP‐4 inhibitor treatment between January 1st, 2015, and August 31st, 2022, were eligible. Drug details are listed in Table S1. The index date was defined as the date of the first prescription of either SGLT2 inhibitors or DPP‐4 inhibitors, with no use in the previous year. We excluded patients with records of type 1 diabetes mellitus or gestational diabetes, acquired immune deficiency syndrome, end‐stage renal diseases, regular dialysis or renal transplantation, and no laboratory test results of estimated glomerular filtration rate (eGFR) or HbA1c in 1 year before the index date. We also excluded patients with less than five measurements of HbA1c in 3 years either before or after the index date, which was indispensable for the measurement of HVS (Table 1 and Figure S1).

TABLE 1.

Baseline characteristics of patients initiating SGLT2 inhibitors or DPP‐4 inhibitors before and after 1:1 propensity score matching.

	Before PSM				After PSM
	DPP‐4 inhibitors	SGLT2 inhibitors	p‐value	SMD	DPP‐4 inhibitors	SGLT2 inhibitors	p‐value	SMD
	(N = 14398)	(N = 5807)			(N = 4612)	(N = 4612)
Calendar year at cohort entry			<0.001	0.687			0.244	0.063
2015	2053 (14.3)	121 (2.1)			83 (1.8)	118 (2.6)
2016	2189 (15.2)	413 (7.1)			355 (7.7)	378 (8.2)
2017	2180 (15.1)	600 (10.3)			533 (11.6)	527 (11.4)
2018	1924 (13.4)	713 (12.3)			609 (13.2)	584 (12.7)
2019	2342 (16.3)	1069 (18.4)			890 (19.3)	845 (18.3)
2020	2015 (14.0)	1360 (23.4)			1064 (23.1)	1070 (23.2)
2021	1478 (10.3)	1372 (23.6)			950 (20.6)	971 (21.1)
2022	217 (1.5)	159 (2.7)			128 (2.8)	119 (2.6)
Age, mean (SD)	68.81 (12.18)	60.44 (11.10)	<0.001	0.718	61.94 (11.71)	61.93 (10.66)	0.995	<0.001
Sex, male	7597 (52.8)	3595 (61.9)	<0.001	0.186	2748 (59.6)	2744 (59.5)	0.949	0.002
Smoking	5391 (37.4)	2451 (42.2)	<0.001	0.097	1863 (40.4)	1873 (40.6)	0.849	0.004
Obesity	8225 (57.1)	4205 (72.4)	<0.001	0.324	3163 (68.6)	3202 (69.4)	0.392	0.018
Diabetes duration			<0.001	0.074			0.819	0.02
<1 years	28 (0.2)	12 (0.2)			10 (0.2)	8 (0.2)
1–5 years	1706 (11.8)	804 (13.8)			660 (14.3)	647 (14.0)
5–10 years	3067 (21.3)	1308 (22.5)			1058 (22.9)	1033 (22.4)
≥10 years	9597 (66.7)	3683 (63.4)			2884 (62.5)	2924 (63.4)
Cerebrovascular disease	2399 (16.7)	629 (10.8)	<0.001	0.17	530 (11.5)	541 (11.7)	0.745	0.007
Congestive heart failure	1364 (9.5)	486 (8.4)	0.015	0.039	347 (7.5)	361 (7.8)	0.611	0.011
Chronic renal failure	1018 (7.1)	155 (2.7)	<0.001	0.206	141 (3.1)	139 (3.0)	0.952	0.003
Hypertension	11076 (76.9)	4376 (75.4)	0.018	0.037	3456 (74.9)	3484 (75.5)	0.515	0.014
Hyperlipidaemia	8378 (58.2)	3910 (67.3)	<0.001	0.19	2992 (64.9)	3014 (65.4)	0.646	0.01
Liver disease	688 (4.8)	355 (6.1)	<0.001	0.059	255 (5.5)	269 (5.8)	0.559	0.013
Ischaemic heart disease	2506 (17.4)	1318 (22.7)	<0.001	0.132	881 (19.1)	940 (20.4)	0.129	0.032
Atrial fibrillation	825 (5.7)	256 (4.4)	<0.001	0.06	203 (4.4)	213 (4.6)	0.652	0.01
Microvascular complications	3120 (21.7)	1223 (21.1)	0.35	0.015	863 (18.7)	914 (19.8)	0.187	0.028
Cancer	1498 (10.4)	433 (7.5)	<0.001	0.104	385 (8.3)	380 (8.2)	0.88	0.004
ACEIs or ARBs	9950 (69.1)	4111 (70.8)	0.019	0.037	3199 (69.4)	3210 (69.6)	0.821	0.005
Diuretics	3556 (24.7)	1177 (20.3)	<0.001	0.106	912 (19.8)	934 (20.3)	0.585	0.012
β‐blockers	5660 (39.3)	2349 (40.5)	0.138	0.023	1779 (38.6)	1805 (39.1)	0.593	0.012
Calcium channel blockers	9199 (63.9)	3247 (55.9)	<0.001	0.163	2634 (57.1)	2649 (57.4)	0.768	0.007
Aspirin	4931 (34.2)	1912 (32.9)	0.075	0.028	1432 (31.0)	1461 (31.7)	0.53	0.014
Statin	10956 (76.1)	4745 (81.7)	<0.001	0.138	3659 (79.3)	3698 (80.2)	0.325	0.021
Other lipid‐lowering drugs	813 (5.6)	489 (8.4)	<0.001	0.109	323 (7.0)	341 (7.4)	0.493	0.015
Insulin	6355 (44.1)	3050 (52.5)	<0.001	0.168	2194 (47.6)	2269 (49.2)	0.123	0.033
Metformin	11734 (81.5)	5526 (95.2)	<0.001	0.436	4370 (94.8)	4340 (94.1)	0.188	0.028
Number of second‐line antidiabetic drugs	0.86 (0.51)	0.96 (0.71)	<0.001	0.173	0.93 (0.52)	0.92 (0.69)	0.66	0.009
HbA1c%, mean (SD)	9.01 (1.70)	9.04 (1.56)	0.17	0.022	8.97 (1.72)	9.01 (1.56)	0.2	0.027
eGFR (mg/min/1.73 m2), mean (SD)	69.34 (33.46)	86.92 (27.11)	<0.001	0.577	85.41 (34.08)	84.74 (26.29)	0.288	0.022
Baseline HVS >80%	4876 (33.9)	1768 (30.4)	<0.001	0.073	1429 (31.0)	1427 (30.9)	0.982	0.001

Abbreviations: ACEI, angiotensin‐converting enzyme inhibitor; ARB, angiotensin receptor blocker; DPP‐4, dipeptidyl peptidase‐4; eGFR, estimated glomerular filtration rate; HbA1c, haemoglobin A1c; HVS, HbA1c variability score; SD, standard deviation; SGLT2, sodium‐glucose cotransporter‐2; SMD, standardized mean difference.

The measurement of HVS followed the method proposed by Angus Forbes et al. ⁴ In brief, HVS for each individual was calculated by counting the number of times successive HbA1c differed by ≥0.5% in a pre‐specified time period, divided by the number of comparisons, and multiplied by 100. Baseline HVS was calculated using all HbA1c results within the 3‐year period prior to cohort entry, with a minimum of 5 HbA1c measurements required to ensure accuracy. HVS indicated the frequency of HbA1c fluctuation instead of the absolute value change. HVS was comparable to the standard deviation and coefficient of variation of HbA1c but more translatable in clinical practice. ⁴ , ⁶ , ²³ The present study utilized a cut‐off value of >60% for high HbA1c variability, which was selected based on prior evidence of association between HVS and increased risk of cardiovascular and renal diseases. ³ , ⁴ , ⁹

2.3. Outcomes

The primary outcome was the post‐treatment HVS measured within 3 years after treatment initiation. The secondary outcomes were (1) MACE, a composite of cardiovascular death, acute myocardial infarction (AMI), and ischaemic or haemorrhagic stroke; (2) serious renal events (SRE), a composite of renal failure (initiation of dialysis, renal transplantation or sustained eGFR <15 mL/min/1.73 m²), a sustained eGFR decline of more than 50% of the baseline and death due to renal diseases ²⁴ ; (3) all‐cause death; (4) cardiovascular death; (5) AMI; (6) ischaemic or haemorrhagic stroke; (7) hospitalization for heart failure (HHF); and (8) hospitalization for renal diseases. ²⁴ Patients were followed until they experienced the event of interest, died, or the end of data collection (December 31st, 2023).

2.4. Covariates

Baseline covariates included demographic characteristics (calendar year at cohort entry, age, sex, diabetes duration), lifestyle factors (smoking, obesity), comorbidities (cerebrovascular disease, congestive heart failure, chronic renal failure, hypertension, hyperlipidemia, liver disease, ischaemic heart disease, atrial fibrillation, diabetic microvascular complications, cancer), glucose‐lowering agents (insulin, metformin, number of second‐line antidiabetic drugs), other medication prescriptions (angiotensin‐converting enzyme inhibitor or angiotensin receptor blocker, diuretics, β‐blockers, calcium channel blockers, aspirin, statin, other lipid‐lowering drugs), and laboratory test results (HbA1c, eGFR and baseline HVS). HbA1c and eGFR were selected as the last recorded values within 1 year before cohort entry. eGFR was calculated based on serum creatinine laboratory results using the Modification of Diet in Renal Disease (MDRD) equation. The definitions used to identify the covariates are listed in Table S1.

2.5. Statistical analysis

Propensity score (PS) matching was performed between SGLT2 inhibitor users and DPP‐4 inhibitor users to diminish confounding effects. The propensity scores were calculated through a logistic regression model with the aforementioned covariates. SGLT2 inhibitor users were 1:1 matched with DPP‐4 inhibitor users using the nearest neighbour approach with a clipper width of 0.02. Covariates with a between‐group standardized mean difference (SMD) <0.1 were considered balanced.

To compare the effectiveness of SGLT2 inhibitors vs DPP‐4 inhibitors in reducing HbA1c variability, we calculated HVS within 3 years after treatment initiation for each individual. HVS was treated as a categorical variable for analysis purposes, with five quintile groups: 0%–20%, 21%–40%, 41%–60%, 61%–80%, and 81%–100%. ³ , ⁹ We performed logistic regression analysis to assess the association between SGLT2/DPP‐4 inhibitors and the post‐treatment HVS. Odds ratios (ORs) with 95% confidence interval (CI) were calculated.

The number of incidences and incidence rates (IRs) per 1000 person‐years for the cardiovascular and renal outcomes were calculated. The Cox proportional hazards regression model was used to estimate hazard ratios (HR) and 95% CIs of SGLT2 inhibitors therapy vs DPP‐4 inhibitors therapy. Cumulative incidence curves for MACE and SRE were plotted based on the Kaplan–Meier method, and log‐rank tests were used to compare hazard rates between therapies, with two‐sided p values <0.05 considered significant.

We performed subgroup analysis to identify potential heterogeneity of treatment effect. Patients were divided into subgroups according to baseline HbA1c (≤7% or >7%), age (≤60 or >60), sex (male or female), smoking history (yes or no), history of cardiovascular diseases (yes or no), history of diabetic microvascular complications (yes or no), and history of obesity (yes or no). Moreover, several sensitivity analyses were performed to validate the robustness of the results. Firstly, we used SD_HbA1c and CV_HbA1c as two alternative measurements of HbA1c variability. Secondly, we used a doubly robust model which adjusted for covariates included in the propensity score model. Thirdly, we used inverse probability of treatment weighting (IPTW) instead of propensity score matching to adjust for confounders. Fourthly, we excluded patients who were concurrently using insulin to minimize the potential impact of insulin on the variability of HbA1c. In the fifth additional analysis, adjustments were made for changes in body mass index (BMI) following treatment initiation, which was restricted to a subgroup of patients with BMI measurements both pre‐ and post‐treatment. Lastly, we adopted cataract hospitalization as a negative control outcome (NCO) to assess the impact of unmeasured confounding factors.

All statistical analyses were performed using R version 4.3.1. Figure graphing was performed using R version 4.3.1 and Microsoft Excel 2021.

3. RESULTS

3.1. Characteristics of study population

A total of 20,205 eligible individuals were identified from CDARS. In the unmatched overall cohort, patients who initiated SGLT2 inhibitors were younger, more frequently men, more likely to have a history of smoking and obesity, and had a higher proportion of insulin and second‐line antidiabetic drug prescriptions. In contrast, patients who initiated DPP‐4 inhibitors had a heavier burden of cardiovascular comorbidities and a lower eGFR at baseline. Following 1:1 PS matching, 9,224 patients were included in the analysis (4,612 pairs). The distributions of baseline characteristics were well‐balanced in the PS‐matched cohort (Figure S2).

3.2. Three years post‐treatment HVS analysis

During the 3 years post‐treatment period, patients initiating SGLT2 inhibitors had a higher representation in the lower quintiles of HVS compared to those initiating DPP‐4 inhibitors (0%–20%: 18.00% vs. 10.84%; 21%–40%: 23.35% vs. 18.52%). When adopting the 0%–20% quintile as a reference, patients initiating SGLT2 inhibitors exhibited a reduced likelihood of being in higher HVS quintiles (21%–40%: OR 0.76, 95% CI 0.66–0.88; 41%–60%: OR 0.57, 95% CI 0.50–0.65; 61%–80%: OR 0.49, 95% CI 0.42–0.56; and 81%–100%: OR 0.40, 95% CI 0.34–0.47) (Table 2 and Figure 1).

TABLE 2.

Analysis on HVS 3 years post‐treatment of patients initiating SGLT2 inhibitors or DPP‐4 inhibitors with baseline HVS >60%.

	SGLT2 inhibitors	DPP‐4 inhibitors	SGLT2 inhibitors vs. DPP‐4 inhibitors
	n (%)	n (%)	OR	95% CI	p‐value
3‐year post‐treatment HVS
0%–20%	830 (18.00)	500 (10.84)	Ref	Ref	–
21%–40%	1077 (23.35)	854 (18.52)	0.76	0.66–0.88	<0.001
41%–60%	1266 (27.45)	1346 (29.18)	0.57	0.50–0.65	<0.001
61%–80%	980 (21.25)	1215 (26.34)	0.49	0.42–0.56	<0.001
81%–100%	459 (9.95)	697 (15.11)	0.40	0.34–0.47	<0.001

Abbreviations: CI, confidence interval; DPP‐4, dipeptidyl peptidase‐4; HVS, HbA1c variability score; OR, odds ratio; SGLT2, sodium‐glucose cotransporter‐2.

FIGURE 1.

Distribution of HVS 3 years post‐treatment in patients initiating SGLT2 inhibitors vs DPP‐4 inhibitors after 1:1 PSM. SGLT2, sodium‐glucose transporter‐2; DPP‐4, dipeptidyl peptidase 4; HVS, HbA1c variability score.

3.3. Cardiovascular and renal outcomes analysis

A total of 345 MACE occurred in patients who received SGLT2 inhibitors and 487 in patients who received DPP‐4 inhibitors in the PS‐matched cohort. The IRs per 1000 person‐years for MACE were 24.72 versus 35.65 in the SGLT2 inhibitor versus DPP‐4 inhibitor group, showing a 31% lower risk (HR 0.69; 95% CI 0.60–0.79). Overall, 478 versus 652 SRE events corresponding to 26.47 versus 37.03 IRs per 1000 person‐years occurred in the SGLT2 inhibitor versus the DPP‐4 inhibitor groups, respectively. SGLT2 inhibitor initiation was associated with a 29% decreased risk of SRE compared with DPP‐4 inhibitors (HR 0.71; 95% CI 0.63–0.80). SGLT2 inhibitors vs DPP‐4 inhibitors initiation was associated with a significantly lower risk of all‐cause death (HR 0.46; 95% CI 0.39–0.53), cardiovascular death (HR 0.58; 95% CI 0.44–0.78), AMI (HR 0.81; 95% CI 0.65–1.01), stroke (HR 0.61; 95% CI 0.51–0.74), HHF (HR 0.70; 95% CI 0.57–0.86), and hospitalization for renal diseases (HR 0.33; 95% CI 0.26–0.40) (Figure 2). Figure S4 illustrates the cumulative incidence of MACE and SRE in patients who received SGLT2 inhibitors vs DPP‐4 inhibitors across subgroups, which was consistent with the results of Cox regression analysis.

FIGURE 2.

Cardiovascular and renal outcomes in patients initiating SGLT2 inhibitors vs DPP‐4 inhibitors after 1:1 PSM. SGLT2, sodium‐glucose transporter‐2; DPP‐4, dipeptidyl peptidase 4; MACE, major adverse cardiovascular events; SRE, serious renal events; AMI, acute myocardial infarction; HHF, hospitalization for heart failure; HR, hazard ratio; CI, confidence interval.

3.4. Subgroup analysis and sensitivity analysis

The association between SGLT2 inhibitors and HVS reduction remained significant after stratifying by baseline HbA1c, age, sex, smoking history, history of cardiovascular diseases, history of diabetic microvascular complications, and history of obesity (Table S2 and Figure S3). In sensitivity analysis, findings remained consistent when SD_HbA1c and CV_HbA1c were adopted instead of HVS as HbA1c variability indicators. Consistent with the primary analysis, the doubly robust model and IPTW model showed significantly reduced post‐treatment HVS associated with SGLT2 inhibitors versus DPP‐4 inhibitors. The results were also consistent when excluding patients with concomitant use of insulin (Table S3). Following adjustment for BMI change, the significant association between SGLT2 inhibitor initiation and post‐treatment HVS persisted. Notably, in the NCO analysis, there was no statistically significant difference in the incidence of cataract hospitalization between initiators of SGLT2 inhibitors and DPP‐4 inhibitors, indicating a null result.

4. DISCUSSION

In this territory‐wide cohort of T2DM patients with high HbA1c variability, our analysis revealed the following observations: (1) SGLT2 inhibitor initiation was associated with greater effectiveness in reducing HbA1c variability compared to DPP‐4 inhibitors; and (2) initiation of SGLT2 inhibitors, as opposed to DPP‐4 inhibitors, was associated with a significantly lower incidence rate of cardiovascular and renal adverse events.

The impact of HbA1c variability on the development of cardiovascular and renal diseases initiates at the subclinical stage, predisposing patients to a heightened risk of cardiovascular and renal events even prior to the appearance of detectable clinical symptoms. ³ , ⁴ , ⁵ , ⁹ Oxidative stress is identified as a pivotal factor in the damage associated with HbA1c variability. ²⁵ , ²⁶ , ²⁷ During glycaemic fluctuations, the mitochondrion generates superoxide, which accelerates the apoptosis of endothelial cells, leading to vascular dysfunction and ultimately contributing to cardiovascular and renal damage. ²⁵ , ²⁸ The compelling evidence linking elevated HbA1c variability to an increased risk of adverse cardiovascular and renal outcomes has driven the focus of this study and underscored the clinical relevance of our findings.

Our study indicated that SGLT2 inhibitors outperformed DPP‐4 inhibitors in reducing HbA1c variability, as evidenced by 68.80% and 58.54% of patients transitioning from HVS ≥60% to HVS <60% after drug initiation, respectively. This finding is consistent with a post‐hoc analysis of the EMPA‐REG OUTCOME trial, which documented a nearly one‐third reduction in SD_HbA1c and a one‐fifth reduction in CV_HbA1c associated with empagliflozin. ²⁹ The superiority of SGLT2 inhibitors in mitigating HbA1c variability can be attributed to their mechanism of action. SGLT2 inhibitors act by inhibiting the SGLT2 protein in the kidneys, leading to increased glucose excretion, with minimal impact on insulin secretion or sensitivity. ³⁰ This mechanism confers a reduced susceptibility to glycaemic fluctuations triggered by changes in insulin levels among users of SGLT2 inhibitors. ²⁶ Furthermore, the extent of glucose excretion induced by SGLT2 inhibitors is contingent upon serum glucose levels. ³¹ In situations where serum glucose levels dip below the normal threshold, urinary glucose excretion is suppressed, preventing further reduction in serum glucose levels. ³² , ³³ This mode of action serves as a safeguard against hypoglycaemia, promoting long‐term glycaemic stability.

Although SGLT2 inhibitors are widely accepted as a cardiorenal protective therapy for diabetes, published RCTs have not sufficiently examined their efficacy in patients with high HbA1c variability. This absence can be attributed to several factors. Firstly, RCTs typically rely on a single measurement of HbA1c at cohort entry as a baseline characteristic rather than tracing back over an extended period. ¹³ , ¹⁴ , ¹⁵ , ¹⁶ Secondly, the stringent inclusion criteria of RCTs often result in the exclusion of many patients with high HbA1c variability. ¹³ , ¹⁴ , ¹⁶ Consequently, existing RCT evidence about the effectiveness of SGLT2 inhibitors in reducing cardiovascular and renal adverse outcomes did not cover the population with high HbA1c variability. Our study demonstrated a 31% lower risk of MACE and a 29% lower risk of SRE compared to DPP‐4 inhibitors in patients with baseline HVS ≥ 60%, highlighting the significant benefit of SGLT2 inhibitors in patients with high HbA1c variability. This study bridges an evidence gap by presenting compelling findings that underline the substantial advantages of SGLT2 inhibitors in patients with frequent glycaemic fluctuations, thereby providing valuable insights into the management of this population.

5. CLINICAL IMPLICATION

Despite the recognition of individuals with high HbA1c variability being at high risk for cardiovascular and renal events, there remains a lack of guideline recommendations concerning the management of this specific population. In our current study, we reported for the first time that SGLT2 inhibitors were superior to DPP‐4 inhibitors in reducing HbA1c variability and demonstrated the effectiveness of SGLT2 inhibitors in preventing cardiovascular and renal adverse outcomes in T2DM patients with high HbA1c variability. These results added compelling evidence for the consistency and robustness of SGLT2 inhibitors efficacy using real‐world data. For individuals diagnosed with T2DM and exhibiting high HbA1c variability, clinical guidelines should prioritize the utilization of SGLT2 inhibitors due to their demonstrated effectiveness in stabilizing HbA1c levels and providing cardiovascular and renal protection.

6. STRENGTHS AND LIMITATIONS

The strength of our study included the utilization of the well‐validated territory‐wide database, CDARS. By utilizing CDARS, we were able to access comprehensive and complete data encompassing the entire population of Hong Kong, thereby enhancing the representativeness and generalizability of our findings. Additionally, we conducted a thorough sensitivity analysis employing different study settings, which further validated our study results.

Our study has several limitations that warrant consideration. Firstly, a notable proportion of patients were excluded due to insufficient HbA1c measurements, potentially introducing selection bias. To mitigate this issue, we conducted a sensitivity analysis using SD_HbA1c and CV_HbA1c as alternative measures of HbA1c variability, which required a minimum of 3 HbA1c measurements, thereby encompassing a larger cohort. While the sensitivity analysis yielded results consistent with the main analysis, it is important to acknowledge that individuals with low health awareness and those who died within 3 years after treatment initiation may be under‐represented in our cohort. Secondly, it is crucial to recognize that patients with high HbA1c variability often present with elevated HbA1c levels. Therefore, the interpretation of findings related to HbA1c variability should consider the potential impact of absolute HbA1c values. Nonetheless, subgroup analyses stratified by HbA1c levels (>7% or ≤7%) produced similar results, suggesting a minimal effect of ambient hyperglycaemia. Thirdly, despite adjusting for a broad array of clinical and laboratory parameters, concerns regarding residual confounding persist. Nevertheless, the NCO analysis yielded a null result, indicating that residual confounding factors were unlikely to substantially bias the relationship between exposure and outcomes. Lastly, it is essential to note that our study was confined to the Hong Kong population, predominantly comprising individuals of Asian descent. Therefore, further investigations are necessary to ascertain the generalizability of our study findings to other ethnic groups.

7. CONCLUSION

In this territory‐wide cohort study involving patients with T2DM and HVS >60%, a subset of vulnerable individuals at high risk of adverse events, which are often under‐represented in RCTs, SGLT2 inhibitors were associated with superior effectiveness over DPP‐4 inhibitors in reducing HbA1c variability. Additionally, the use of SGLT2 inhibitors, compared to DPP‐4 inhibitors, was linked to a significantly decreased risk of cardiovascular and renal adverse events. In summary, SGLT2 inhibitors emerge as an attractive therapeutic choice for individuals with T2DM displaying high HbA1c variability.

AUTHOR CONTRIBUTIONS

R.G. and M.‐Z.W. completed the initial data preparation and statistical analyses. R.G. and K.‐H.Y. drafted the manuscript. A.P., C.C., and A.‐P.C. were involved with the conception of the study. Q.‐W.R., J.‐Y.H., J.‐N.Z., W.‐L.G., W.O., J.T., T.‐H.K.T., C.T.‐W.T., C.‐Y.Z., H.‐C.X., Y.‐X.L., and Y.‐M.H. contributed to the statistical analyses. C.S.P.L. and K.‐H.Y. provided the clinical expertise. All authors critically reviewed and contributed to the intellectual content of the manuscript and approved the final version of the manuscript. R.G. and K.‐H.Y. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

FUNDING INFORMATION

This study was supported by the Sanming Project of Medicine in Shenzhen, China (No. SZSM202411021), the HKU‐SZH Fund for Shenzhen Key Medical Discipline (No. SZXK2020081), the National Natural Science Foundation of China (No. 82270400) and the Natural Science Foundation of Guangdong Province (No. 2023A1515010731).

CONFLICT OF INTEREST STATEMENT

Carolyn SP Lam is supported by a Clinician Scientist Award from the National Medical Research Council of Singapore; has Received research support from Novo Nordisk and Roche Diagnostics; has Served as consultant or on the Advisory Board/Steering Committee/Executive Committee for Alnylam Pharma, AnaCardio AB, Applied Therapeutics, AstraZeneca, Bayer, Biopeutics, Boehringer Ingelheim, Boston Scientific, Bristol Myers Squibb, Corteria, CPC Clinical Research, Cytokinetics, Eli Lilly, Impulse Dynamics, Intellia Therapeutics, Ionis Pharmaceutical, Janssen Research & Development LLC, Medscape/WebMD Global LLC, Merck, Novartis, Novo Nordisk, Quidel Corporation, Radcliffe Group Ltd., Roche and Us2.ai; and serves as Co‐founder & non‐executive director of Us2.ai. Other co‐authors have no conflict of interest to disclose.

ETHICS STATEMENT

The study was approved by the institutional review board of the University of Hong Kong and the West Cluster of the Hong Kong Hospital Authority. Informed patient consent was not required as the data used in this study are anonymized.

CONSENT FOR PUBLICATION

All authors have approved the final manuscript for publication.

Supporting information

Table S1. Covariates definitions.

Table S2. Subgroup analysis stratified by baseline HbA1c, age, sex, history of smoking, history of cardiovascular diseases, history of diabetic microvascular complications.

Table S3. Sensitivity analysis for the primary outcome.

Figure S1. Study flowchart.

Figure S2. Covariates were well balanced between treatment groups after PSM.

Figure S3. Distribution of three years post‐treatment HVS across subgroups.

Figure S4. Cumulative incidence of composite cardiovascular and renal outcomes by treatment group.

Guo R, Pandey A, Chandramouli C, et al. The effect of sodium‐glucose cotransporter 2 inhibitors on HbA1c variability and cardiovascular and renal adverse outcome in patients with T2DM . Diabetes Obes Metab. 2025;27(9):4720‐4728. doi: 10.1111/dom.16509

DATA AVAILABILITY STATEMENT

Data are available upon reasonable request by contacting Prof. Yiu Kai‐Hang.