The effects of non-insulin anti-diabetic medications on the diabetic microvascular complications: a systematic review and meta-analysis of randomized clinical trials

Abstract

Introduction

Although the onset and progression of diabetic microvascular complications are linked to glycemic control, various antihyperglycemic drugs with distinct treatment targets may positively impact microvascular lesions beyond their glucose-lowering effects. Therefore, this systematic review emphasizes the clinical therapeutic implications of non-insulin anti-diabetic medications for diabetic microvascular complications.

Methods

We retrieved published literature reporting randomized clinical trials (RCTs) on the effects of microvascular complications, including diabetic nephropathy (DN), diabetic peripheral neuropathy (DPN), and diabetic retinopathy (DR), from authenticated clinical databases: PubMed, Excerpta Medica database (EMBASE), and Web of Science. We synthesized data, including the continuous variable indices: estimated glomerular filtration rate (eGFR), urinary albumin to creatinine ratio (UACR), and urinary albumin excretion rate (UAE). Indices measuring cardiovascular autonomic neuropathy (CAN), vibration detection threshold (VDT), and retinal nerve fiber thickness (RNFL) were used to calculate microvascular effects. We also synthesized dichotomous variable indices, including the risks for DR and DPN.

Results

According to our analyses, there was sparse evidence strongly supporting that metformin (MET), Sulfonylurea (SUs), Repaglinide (Repa), or α-Glucosidase inhibitors (α-GIs) could benefit diabetic microvascular complications when adopted as monotherapy. Regardless of the no change in eGFR, two trials reporting Thiazolidinediones (TZDs) significantly reduced the UACR, while other clinical trials reported an increase in VDT and improvement in DR. Sodium glucose co-transporter inhibitors (SGLT-2i) and Glucagon-like peptide-1 receptor agonists (GLP-1RA) both showed protective effects in preventing eGFR decline, with only SGLT-2i demonstrating a significant reduction in UACR. A recent trial showed that Dipeptidyl Peptidase IV inhibitors (DPP-IVi) may potentially reduce the risk of DPN, while GLP-1RA did not prove to alter the measures of CAN and DPN. However, the SUSTAIN 6 trial revealed that Semaglutide may increase the risk of DR.

Conclusion

Besides their anti-hyperglycemic properties, some currently reviewed medications may exhibit unique anti-microvascular abilities. Due to ambiguous and conflicting available results, more emerging or ongoing trials will address this issue and could benefit clinical strategies for personalized treatment practices.

Clinical trial number

Not applicable.

Key summary points

1. While diabetic microvascular complications can be prevented through intensive glycemic control, various anti-diabetic medications may exhibit differing levels of microvascular protective abilities due to specific class effects.

2. According to the current data analysis, there is insufficient evidence demonstrating that traditional anti-diabetic drugs prevent diabetic nephropathy (DN) and diabetic retinopathy (DR); there is even scant literature reporting their effectiveness in diabetic peripheral neuropathy (DPN).

3. Novel anti-diabetic medications, particularly SGLT-2 inhibitors, have shown benefits in improving diabetic nephropathy (DN), especially regarding the enhancement of urinary albumin-to-creatinine ratio (UACR). Data suggests that DPP-IV inhibitors may improve diabetic peripheral neuropathy (DPN). However, there exists heterogeneity in the data supporting the effects of GLP-1 receptor agonists (GLP-1 RA) on UACR. The improvement in DN appears to be independent of any enhancements in proteinuria. In the Sustain 6 study of semaglutide, the data does not seem favorable regarding the risk of diabetic retinopathy (DR), which contrasts with previous results and findings related to other GLP-1 medications, such as liraglutide.

Introduction

Type 2 diabetes mellitus (T2D) is the most common form of diabetes, characterized by high insulin resistance and varying degrees of relative insulin deficiency [1]. Diabetic microvascular complications are serious chronic consequences of prolonged hyperglycemia, primarily affecting small blood vessels and manifesting [2]. These conditions result from high glucose-induced endothelial dysfunction, oxidative stress, inflammation, and the accumulation of advanced glycation end-products (AGEs), which together lead to basement membrane thickening, increased vascular permeability, and tissue hypoxia [3]. Diabetic microvascular complications present significant health risks due to their effects on small blood vessels, resulting in severe and potentially debilitating outcomes [4]. These complications primarily encompass diabetic retinopathy (DR), diabetic nephropathy (DN), and diabetic peripheral neuropathy (DPN), each having distinct but overlapping harms [5].

Diabetic Nephropathy (DN): DN affects the renal microvasculature, resulting in proteinuria, a decline in glomerular filtration rate, and ultimately, end-stage renal disease (ESRD). ESRD necessitates dialysis or transplantation, both of which pose high morbidity and mortality risks. Additionally, DN exacerbates cardiovascular disease, increasing overall mortality risk. It accounts for a substantial proportion of chronic kidney disease cases worldwide [6]. Diabetic Retinopathy (DR): DR damages the microvasculature of the retina, leading to vision impairment and, in severe cases, blindness. It progresses from non-proliferative stages (microaneurysms, hemorrhages) to proliferative stages (neovascularization), which increases the risk of retinal detachment and vitreous hemorrhage. Globally, DR is a leading cause of blindness among working-age adults, significantly reducing quality of life and rising healthcare costs [7]. Diabetic Peripheral Neuropathy (DPN): DPN damages peripheral nerves, resulting in sensory loss, neuropathic pain, and motor dysfunction. This heightens the risk of foot ulcers and infections, which can lead to amputations. DPN also impairs balance, raising the risk of falls and fractures, particularly in elderly patients. The chronic pain and disability associated with DPN severely impact mental health and daily functioning [8].

In recent years, significant progress has been made in the study of non-insulin antidiabetic medications for ameliorating microvascular complications, with various drugs demonstrating protective effects through distinct mechanisms. Previous research has shown that intensive glycemic control or hypoglycemic treatment can reduce or delay the risk or occurrence of microvascular complications [9, 10]. However, due to the action on different pathophysiological targets of T2D by other drugs, and the resulting divergence in the mechanisms of these discrete medications, the effects of non-insulin antidiabetic medicines on these complications may vary [11]. Additionally, appropriate anti-diabetic treatment is essential to control hyperglycemia and reduce the incidence of microvascular disease, thereby delaying the onset of complications.

In clinical practice, it is crucial to select medications based on the functional status of the organ and pathophysiology. Although most drugs have been evaluated for macrovascular safety, the precise role of individual medications in preventing complications is less clear. Therefore, it is essential to compare the properties of drugs across various complications. This research focused on non-insulin antidiabetic medications impacting major microvascular diseases, such as DN, DPN, and DR. We evaluated the effects of eight non-insulin antidiabetic drugs on T2D microvascular complications, including metformin (MET), sulfonylureas (SUs), repaglinide (Repa), α-glucosidase inhibitors (α-GIs), thiazolidinediones (TZDs), sodium-glucose co-transporter inhibitors (SGLT-2i), dipeptidyl peptidase-IV inhibitors (DPP-IVi), and glucagon-like peptide one receptor agonists (GLP-1RAs).

Methods

Criteria for selecting studies for this review

This study was a systematic meta-analysis of the current literature, and the entire method adhered to PRISMA guidelines.

Types of studies

We included completed randomized controlled trials (RCTs) on the effects of diabetic microvascular complications, such as diabetic nephropathy (DN), diabetic peripheral neuropathy (DPN), and diabetic retinopathy (DR).

Types of participants

Adult patients diagnosed with T2D, including those who have diabetic microvascular complications.

Types of interventions

Interventions were classified based on the currently available classes of standard drugs: MET, SUs, Repa, α-GIs, TZDs, SGLT-2 inhibitors, DPP-IV inhibitors, and GLP-1 receptor agonists. Comparatives were chosen, including placebo or MET, SUs, as indicated by specific randomized controlled trials (RCTs).

Types of outcome measures

We analyzed representative indices calculated in DN, DPN, and DR, including continuous variables such as estimated glomerular filtration rate (eGFR) and urinary albumin to creatinine ratio (UACR) or urinary albumin excretion rate (UAE) across all studies reporting results on renal microvascular complications. We also included continuous variable indices of measures for cardiovascular autonomic neuropathy (CAN), vibration detection threshold (VDT), as well as a dichotomous variable index of risk for DPN. Additionally, we assessed continuous variables of retinal nerve fiber thickness (RNFL) and a dichotomous variable index of risk for DR. The primary outcome measure encompassed indices commonly used in renal microvascular outcomes, including eGFR, UACR, and UAE. In the cases of DPN and DR, the risk of microvascular disease was identified as the primary outcome measure.

Primary outcomes

For DN, we compared the authors’ data on the decline in eGFR, UACR, and UAE. We extracted binary variables as risk indices for DPN, and our primary outcome included binary variables as risk indices for DR.

Secondary outcomes

For DPN, we extracted continuous indices calculated in CAN and VDT; for DR, our outcome of interest included continuous variables of RNFL.

Search methods for the identification of studies

Electronic searches

We searched the literature in clinical databases (886 pieces of literature were found), including PubMed (282 pieces found), the Excerpta Medica database (EMBASE) (315 pieces found), and the Web of Science (289 pieces found). The publication dates ranged from December 1983 to July 2022.

We utilized the strategy of Medical Subject Headings (MESH) combined with entry terms, following the principles of “patients, interventions, comparisons, outcomes, and study design” (PICOS) in evidence-based medicine searching on PubMed, using the strings (see Supplemental materials).

We adopted a similar approach to “PICOS,” which was used in the PubMed database search to explore the entire literature across other databases. No restrictions were placed on the searches regarding language or publication status.

The eligibility and selection criteria were limited to clinical studies, including RCT, with reported outcomes on T2DM within the current defined time range, excluding meta-analyses, case series, mechanistic studies, conference abstracts, animal experiments, in-vitro studies, and non-T2DM conditions, such as T1D and other types of DM. The selection details are illustrated in Fig. 1.

Fig. 1.

Is the study flow diagram

Searching other resources

We examined the reference lists of the studies included in this research and reviewed the reference lists of all trials identified by the aforementioned techniques.

Data collection and analysis

Selection of studies

Initially, we used Endnote’s literature management software (version X8) while excluding duplicated literature. The two review authors (YY and YYL) then independently screened the literature appropriate for our study. Literature excluded included review articles, systematic or meta-analysis articles, animal studies, basic studies, meeting abstracts, and case reports. Afterward, we screened the titles and abstracts of studies for potentially relevant content. Discussions helped resolve any disagreements regarding study selection. We obtained the full texts suitable for potential data extraction and analysis and assessed their appropriateness for inclusion in the review. For any divergences, a third review author (LJC) was consulted. We documented the selection process in sufficient detail to complete a PRISMA flow diagram (Fig. 1).

This PRIMS study flow diagram outlines and displays the details of the literature selection flow for the current clinical investigation are as follows: (1) exclusion of duplicated literature; (2) title and abstract screening, which excludes inappropriate literature; (3) assessment of full-text articles for eligibility, which involves content verification to select relevant themes and report outcomes following the current study aims; (4) qualitative synthesis of outcomes, selecting studies where outcome measures align with the evidence synthesis; (5) quantitative synthesis, screening to include studies without obvious bias based on the data; (6) synthesis of the evidence and data appropriate for the current study.

Data extraction and management

Data extraction and study characteristics from all included studies were conducted independently by two review authors (YY and YYL), with any disagreements discussed. A third review author (LJC) was consulted as needed. Subsequently, the results were verified by a fourth review author (SW). We utilized a form to collect data on study characteristics and outcomes. The study characteristics extracted from the included content were as follows:

• Methods: study design, intervention duration, follow-up period length, study location, study setting, withdrawals, and study date.

• Participant characteristics include number, mean age, age range, gender, diagnosis, condition severity, diagnostic criteria, laboratory data, and inclusion and exclusion criteria.

• Interventions: intervention, comparison, and concurrent medications.

• Outcomes: outcomes specified and collected at baseline and at the time of intervention completion, along with follow-up measures reported at any other time point.

• Notes: funding for studies and significant conflicts of interest of trial authors.

Assessment of risk of bias in included studies

We utilized the initial version of Cochrane’s ‘Risk-of-bias’ tool to evaluate:

• Random sequence generation.

• Allocation concealment.

• Blinding of participants and personnel.

• Blinding of outcome assessment.

• Incomplete outcome data.

• Selective outcome reporting.

• Other bias.

Each domain was assigned a level of ‘high,’ ‘low,’ or ‘unclear risk of bias,’ along with a supporting quotation from the study report and a justification for our judgment in the ‘Risk-of-bias’ table.

Measures of treatment effect

We analyzed continuous outcomes by calculating the mean difference (MD) with 95% confidence intervals (CI) when studies utilized the same outcome measure, or the standardized mean difference (SMD) with 95% CI when different outcome measures were used. We calculated the risk ratio (RR) for dichotomous outcomes. Effect sizes and 95% confidence intervals (CIs) were computed and presented as forest plots for each treatment comparison.

Unit of analysis issues

The unit of analysis included participants with type 2 diabetes (T2D). In trials with more than two arms, we compared drug classes with the comparators.

Dealing with missing data

In cases of missing data, we made several attempts to resolve the issue, including contacting the corresponding authors of the studies as needed. When data were reported in a graph rather than in a table or text, we estimated the means and standard deviations (SD) using the WebPlotDigitizer tool (https://automeris.io/WebPlotDigitizer). If SD was not displayed, we assessed it from the CI or other measures of variance when possible. If the standard deviation (SD) for follow-up measurements was absent, we used the SD measured at baseline for subsequent follow-up measurements. SD was available for all included studies, and no missing data were identified.

Assessment of heterogeneity

We subsequently performed a random-effects meta-analysis model to account for differences in the patient populations of the included studies and the application of the intervention. In the meta-analysis, we visually inspected forest plots to identify heterogeneity. This process involves evaluating the degree of overlap in confidence intervals, the spread and alignment of point estimates, and the consistency of effect sizes and directions across studies. Observations of outliers or uneven distributions around the pooled effect serve as indicators of potential heterogeneity. These findings are complemented by statistical measures, such as I² or Cochran’s Q test, for a more robust assessment. We evaluated statistical heterogeneity using the Chi² test and the I² statistic. An I² value exceeding 50% was suspected to indicate overt heterogeneity, while values of 75–100% suggested considerable heterogeneity.

Assessment of reporting biases

We aimed to identify the clinical trial registration record and published protocol for each included study. We compared the methods to those in the final published report, when available. If the trial registration or protocol was not available, this was noted in the characteristics of included studies table and assessed as an unclear risk of bias.

Data synthesis

Data were processed as stated in the Cochrane Handbook for Systematic Reviews of Interventions. A formulation form was adopted to address the missing data. The data synthesis and analysis were conducted using Review Manager 5.0 software and checked for heterogeneity.

Results

Description of studies

Results of the search

Our search strategy identified 886 references from electronic databases (Fig. 1). After removing duplicates, 488 unique articles were screened for potential relevance based on title and abstract. We retrieved the full texts of 54 potentially relevant references and screened them for inclusion in the review.

Included studies

For the DN, we compiled literature related to a decline in eGFR and improvements in Proteinuria. For the DPN, we collected literature data. For the DR, we retrieved literature.

Excluded studies

Two hundred sixty-one records were excluded for the following reasons: in vitro study (n = 7), review and meta-analyses (n = 190), conference abstract (n = 1), animal studies (n = 9), guidelines or consensus (n = 5), proceedings (n = 3), commentary (n = 6), study designs or protocols (n = 17), book chapters (n = 2), case reports (n = 1), and others (n = 20). Thirty-two full-text articles were excluded for the following reasons: unrelated themes; lacking full text; T1D; prediabetes; insulin treatment; mechanistic study; other unspecified outcome endpoints; prediction models; prevalence studies; ongoing trials with no reported results. Furthermore, 11 studies were screened out due to a high suspicion of bias, which could influence the analysis results [1].

Risk-of-bias in included studies

Refer to Fig. 2 for the ‘Risk-of-bias’ graph illustrating the proportion of studies with each judgment, and see Fig. 3 for the ‘Risk-of-bias’ summary, which displays all the decisions in a cross-tabulation by trial.

Fig. 2.

Risk-of-bias graph: review authors’ judgments about each risk-of-bias item presented as percentages across all included studies

Fig. 3.

Summary of risk of bias: judgments by review authors regarding each risk of bias item for every included study

Allocation (selection bias)

Only seven trials failed to describe the randomization procedure and were rated as high risk; three were evaluated as unclear risk; eight did not report a concealed allocation procedure and were deemed high risk, while five were assessed as unclear risk (Fig. 3).

Blinding (performance bias and detection bias)

Out of the 54 trials, seven were at risk of blinding participants and personnel, while five trials had unclear risk assessments. Regarding the risk of blinding in outcome assessment, six trials were evaluated as having high risks, and six were deemed unclear (Fig. 3).

Incomplete outcome data (attrition bias)

Though reported losses to follow-up and dropout were relatively low, three trials were unclear in their reporting of dropout rates and in handling incomplete outcome data in most clinical trials.

Selective reporting (reporting bias)

Most clinical trials lacked explicit descriptions of data concerning selective reporting.

Other potential sources of bias

Other potential sources of bias were not evident in each trial.

Effects of interventions

Effects of MET on DN, DPN

In Fig. 4A and B, we present the results of the analyses of MET on eGFR and UAE.

Fig. 4.

presents a forest plot comparing the effects of metformin versus glibenclamide on eGFR (A) and UAE (B); MET plus insulin versus insulin plus placebo on CAN (C) and VDT (D) in T2D

Compared to glibenclamide (Fig. 4A and B), MET showed a significant relative decline in eGFR (MD: -19.00, 95% CI [-36.32, -1.68], p = 0.03), while the UAE did not decrease substantially (MD: -44.00, 95% CI [-195.31, 107.31], p = 0.57) [12]. In Fig. 4C and D, we present the results of the analyses of MET on CAN and VDT, respectively. Compared to controls (MET + Insulin vs. Placebo + Insulin), MET did not significantly alter the CAN (p > 0.05). However, the VDT significantly increased (MD: 4.5, 95% CI [2.31, 6.69], p < 0.0001) [13] Table 1.

Table 1.

The study design and demographic characteristics of clinical trials of MET

Author (Year)	Country	Study Design	Sample Size	Mean Age(years)	% Male	Intervention	Control /Comparator	Outcome Measure	Follow-up Duration (Wks)	References
Amador-Licona, N. (2000)	Mexico; Single center	RCT:	51	48	55	MET 2550 mg	Glibenclamide 20 mg	eGFR; UAE	12	[12]
Hansen, C. S.(2020)	Multicenter	RCT	412	60	68	MET + Inuslin Vs Insulin	Insulin + lacebo	CAN; VDT;	208	[13]

Effects of SUs on DN and DR

In Fig. 5A, B, and C, we present the results of analyses of SUs on eGFR, UACR, and UAE, respectively. Compared to control TZD, glyburide did not significantly alter the eGFR (MD: 0.00, 95% CI [-1.22, 1.22], p = 1.0). Additionally, it did not substantially decrease UACR (MD: -0.78, 95% CI [-1.58, 0.03], p = 0.06) [14], whereas the UAE of glimepiride in another trial significantly improved compared to MET (MD: -70.79, 95% CI [-73.64, -67.94], p < 0.0001) [15]. In Fig. 5D, compared to controls, SUs did not significantly alter the risk of DR (RR: 1.07, 95% CI [0.08, 13.87], p = 0.96, I2 = 77%) [16]. Among the trials, one trial supported the idea that SUs decreased the risk of DR [1]. In contrast, another trial did not find a significant improvement in risk prevention (Fig. 5D Gliclazide vs. Placebo, other SUs vs. Placebo) Table 2.

Fig. 5.

Forest plot comparing the effects of Glyburide vs. TZD on eGFR (A), UACR (B); Glimepiride vs. MET on UAE (C); Gliclazide and other SUs vs. control, and Gliclazide vs. Glibenclamide on the risk of DR (D) with funnel plot (E)in T2D

Table 2.

The study design and demographic characteristics of clinical trials of SUs

Author (Year)	Country	Study Design	Sample Size	Mean Age(years)	% Male	Intervention	Control /Comparator	Outcome Measure	Follow-up Duration (Wks)	References
Lachin, J. M. (2011)	US	RCT:	4351	56	58	glyburide	Rosiglitazone; MET	eGFR; UACR	260	[14]
Mujeeb, M. M. A. (2015)	India	RCT	300	NA	NA	Glimepiride 2 mg qd + losartan 100 mg qd	MET 500 mg bid; Repaglinide 2 mg Q8H + losartan 100 mg qd	UAE	24	[15].
Akanuma, Y. (1988)	Japan	RCT	155	57	52	Gliclazide: 40 mg qid up to 240 mg qid	other sulfonylureas; Diet alone;	DR: funduscopic examinations	260	[16]
Baba, S. 1983	Japan	RCT	289	55	48	Gliclazide:40 mg up to 4 tablets per day	Glibenclamide 2.5 mg up to 4 tablets per day	Funduscopic	24	[17]

Effects of Repa on UAE of DN

In Fig. 6, we present the results of the analyses of Repa in UAE. Compared to the controls MET, Repa significantly decreased the UAE (MD: -70.55, 95% CI [-73.42, 67.68], p < 0.0001) [15] Table 3.

Fig. 6.

displays a forest plot comparing the effects of Repa + MET + losartan versus MET + losartan on UAE in T2D

Table 3.

The study design and demographic characteristics of clinical trials of Repa

Author (Year)	Country	Study Design	Sample Size	Mean Age (years)	% Male	Intervention	Control /Comparator	Outcome Measure	Follow-up Duration (Wks)	References
Mujeeb, M. M. A. (2015)	India	RCT	300	NA	NA	Repa 2 mg Q8H + losartan 100 mg qd	MET; 500 mg bid +losartan	UAE	24	[15]

Effects of α-GIs on UACR of DN

Figure 7 illustrates the results of analyses indicating that α-GIs affect UACR. In comparison to control MET, α-GIs significantly reduced the UACR (MD: -2.07, 95%CI [-3.13, -1.01], p = 0.0001) [18] Table 4.

Fig. 7.

shows the forest plot comparing the effects of α-GIs and MET on UACR in T2D

Table 4.

The study design and demographic characteristics of clinical trials of α-GI

Author (Year)	Country	Study Design	Sample Size	Mean Age(years)	% Male	Intervention	Control /Comparator	Outcome Measure	Follow-up Duration (Wks)	References
Song, L. L., (2021)	China	RCT	352	50	58	Acarbose 100 mg TID	MET:1500 mg QD	UACR	48	[18]

Effects of TZDs on DN

In Figs. 8A, B, and C, the analysis results demonstrate the effect of TZDs on eGFR compared to placebo [19], SU (glipizide, glyburide, and glimepiride) [14, 20, 21], and other unspecified active comparators [22], respectively. The UACR results were compared to those of a placebo (D) [23], SU (glyburide and glimepiride) (E), and other unspecified active comparators (F), while the UAE results were shown for TZD compared to SU (G). When compared to controls, TZDs did not have a significant impact on eGFR (p > 0.05). The UACR significantly decreased compared to placebo (MD: -13.00, 95% CI [-14.75, -11.25], p < 0.00001). The UAE significantly increased when compared to SU (glibenclamide) in two studies (MD: -80.09, 95% CI [-125.42, -34.77], p = 0.0005, I2 = 85%) [24, 25] Table 5.

Fig. 8.

Forest plot comparing the effect of TZDs on eGFR versus placebo (A), sulfonylurea (B) with funnel plot (H); and other unspecified active comparators (C). The effect of TZDs on UACR is presented in comparison to placebo (D), sulfonylurea with (E) funnel plot (I); and other unspecified active comparators (F), along with UAE compared to sulfonylurea (G) with funnel plot (J) in T2D

Table 5.

The study design and demographic characteristics of clinical trials of TZD

Author (Year)	Country	Study Design	Sample Size	Mean Age(years)	% Male	Intervention	Control/Comparator	Outcome Measure	Follow-up Duration (Wks)	References
Pistrosch, F.(2012)	Germany	RCT	28	66	92	Rosiglitazone 4 mg bid	Placebo	eGFR;	52	[19]
Agarwal, R. 2005	US	RCT	44	65	22	pioglitazone	glipizide	eGFR;	16	[20]
Lachin, J. M. (2011)	US	RCT:	4351	56	58	Rosiglitazone;	Metformin; glyburide	eGFR; UACR	260	[14]
Petrica, L. (2009)	Romania	RCT	34	63	41	Rosiglitazone	glimepiride	eGFR; UACR	52	[21]
Ho, C. C. (2022)	Taiwan, China	Retrospective cohort	742	65	58	pioglitazone	non-pioglitazone	eGFR; UACR	364	[22],
Miyazaki, Y. (2007)	US	RCT	29	55	55	Rosiglitazone	Placebo	UACR	12	[23]
Nakamura, T. (2001)	Japan	RCT	58	52	65	Pioglitazone 30 mg QD	Placebo	UAE	24	[24]
Nakamura, T. (2006)	Japan	RCT	68	56	56	Pioglitazone 30 mg QD	Glibenclamide 5 mg QD; voglibose 0.6 mg QD; Nateglinide 270 mg/d	UAE	552	[25]

Effects of TZDs on DPN and DR

Figure 9A shows the results of TZDs on VDT of DPN. Compared to controls, TZDs combined with exenatide significantly increased VDT when compared to basal-bolus insulin therapy (MD: 4.40, 95% CI [0.52, 8.28], p = 0.03) [26]. Figure 9B presents the results relative to controls (no TZD, all other non-insulin anti-diabetic drugs excluding TZDs); TZDs did not significantly increase the risk of DR (RR: 1.16, 95% CI [0.87, 1.55], p = 0.32) [27] Table 6.

Fig. 9.

Forest plot comparing the effects of TZDs + Exenatide versus basal insulin + bolus insulin on VDT (A) and TZDs versus non-TZDs on DR (B) in T2D

Table 6.

The study design and demographic characteristics of clinical trials of TZD on DPN and DR

Author (Year)	Country	Study Design	Sample Size	Mean Age(years)	% Male	Intervention	Control /Comparator	Outcome Measure	Follow-up Duration (Wks)	References
Ponirakis, G. (2020)	Qatar	RCT	56	52	64	Pioglitazone:30 mg QD Exenatide:2 mg/week	glargine with aspart insulin	corneal confocal microscopy (CCM)	52	[26]
Gower, E. W. (2018)	US	RCT	2856	62	63	TZD	Non-TZD	stereoscopic fundus photographs	208	[27]

Effects of SGLT-2i on DN

Figures 10 and 11 illustrate the results of SGLT-2i on eGFR and UACR，respectively. Compared to placebo, SGLT-2i significantly improved eGFR (MD: 1.38, 95% CI [0.85, 1.91], p < 0.00001, I2 = 98%). Additionally, UACR in SGLT-2i was substantially reduced (MD: -14.62, 95% CI: [-21.28, -7.96], p < 0.0001, I2 = 93%) (Fig. 11) [28–49] Table 7.

Fig. 10.

Forest plot comparing the effect of SGLT-2 inhibitors versus placebo on eGFR in type 2 diabetes (T2D) (A) and funnel plot (B)

Fig. 11.

Forest plot comparing the effect of SGLT-2i versus placebo on UACR with funnel plot in T2D

Table 7.

The study design and demographic characteristics of clinical trials of SGLT-2i on DN

Author (Year)	Country	Study Design	Sample Size	Mean Age (years)	% Male	Intervention	Control/Comparator	Outcome Measure	Follow-up Duration (Wks)	Refer-ences
Yale, J. F. (2014)	Multicenter	RCT	269	68	60	Canagliflozin: 100,300 mg QD	Placebo	eGFR, UACR	52	[28]
Heerspink, H. J. (2017)	Multicenter	RCT	1450	56	52	Canagliflozin 100 mg;300 mg	glimepiride	eGFR	104	[29]
Perkovic, V. (2019)	Multicenter	RCT	4401	63	66.1	Canagliflozin 100 mg	Placebo	UACR	126	[30]
Bakris, G. (2020)	Multicenter	RCT	174	65	61	Canagliflozin 100 mg	placebo	eGFR; UACR	130	[31]
Jardine, M. J. (2020)	Mulitcenter	RCT	4401	63	66	Canagligflozin 100 mg	Placebo	eGFR	137	[32]
Li, J. (2020)	Multicenter	RCT	4159	> 30 for ASCVD > 50 for risks of ASCVD	64	Canagliflozin 100 mg up to 300 mg	Placebo	UACR	124	[33]
Lam, D. (2022)	Multicenter	RCT	4330	64	68	Canagliflozin 100 mg;300 mg	Placebo	eGFR, UACR	312	[34]
Wada, T. (2022)	Multicenter	RCT	4401	63	65.2	Canagliflozin 100 mg	Placebo	eGFR, UACR	182	[35]
Dekkers, C. C. J. (2018)	Multicenter	pooled analysis of RCT	220	66	55	Dapagliflozin:5 mg;10 mg	Placebo	eGFR; UACR	104	[36]
Pollock, C. (2019)	Multicenter	RCT	1187	64	70	Dapagliflozin 10 mg Dapagliflozin + Saxagliptin 2.5 mg	Placebo Saxagliptin 2.5 mg	UACR	24	[37]
Mosenzon, O. (2019)	Multicenter	RCT	17,160	65	63	Dapagliflozin 10 mg	Placebo	eGFR, UACR	219	[38]
Chertow, G. M. (2021)	Multicenter	RCT	626	61	67	Dapagliflozin 10 mg	Placebo	eGFR; UACR	172	[39]
Heerspink, H. J. L. (2021)	Multicenter	RCT	4304	62	66	Dapagliflozin 10 mg	Placebo	eGFR, UACR	120	[40]
Mosenzon, O. (2021)	Multicenter	RCT	17,160	age ≥ 55 for male ≥ 60 for female eASCVD ≥ 40	63	Dapagliflozin 10 mg	Placebo	UACR	192	[41]
van Ruiten, C. C. (2021)	Multicenter	RCT	66	64	72.7	Dapagliflozin + Exenatide 10 mg	Placebo	eGFR, UACR	16	[42]
Jongs, N. (2021)	Multicenter	RCT	4304	62	66	Dapagliflozin 10 mg	Placebo	eGFR, UACR	164	[43]
Levin, A. (2020)	Multicenter	RCT	6952	65	72	Empagliflozin 10 mg; 25 mg	Placebo	eGFR, UACR	157	[44]
Kadowaki, T. (2019)	Multicenter	RCT	7020	62	73	Empagliflozin 10 mg;25 mg	Placebo	eGFR	164	[45]
Ruggenenti, P. (2022)	Multicenter	RCT	112	62	70	Empagliflozin: 10 mg; 25 mg	Placebo	eGFR, UACR	120	[46]
Wanner, C. (2020)	Multicenter	RCT	6952	64	72	Empagliflozin: 10 mg, 25 mg QD	Placebo	eGFR, UACR	162	[47]
Cherney, D. Z. I. (2021)	Multicenter	RCT	8246	64	70	Ertugliflozin 5 mg; 15 mg	Placebo	eGFR; UACR	168	[48]
Cherney, D. Z. I. (2021)	Multicenter	RCT	8243	64	70	Ertugliflozin 5 mg; 15mg260	placebo	eGFR; UACR	260	[49]

Effects of SGLT-2i on DPN

Figure 12 presents the results of analyses compared with controls; Cana did not significantly increase the risk of DPN (p > 0.05) [50] Table 8.

Fig. 12.

Forest plot comparing the effect of Canagliflozin versus placebo on the risk of DPN in T2D

Table 8.

The study design and demographic characteristics of clinical trials of SGLT-2i on DPN

Author (Year)	Country	Study Design	Sample Size	Mean Age(years)	% Male	Intervention	Control /Comparator	Outcome Measure	Follow-up Duration (Wks)	References
Liao, J. (2022)	Multicenter	RCT	4401	63	66	Canagliflozin 100 mg	Placebo	Neuropathy events: peripheral /autonomic neuropathy event	118	[50]

Effects of DPP-IV i on DN, DR, and DPN

Figure 13A and B showed the results compared to placebo. DPP-IV i did not significantly decrease the eGFR (MD: -0.63, 95% CI [-2.73, 1.47], p = 0.56, I2 = 99%) [51–55]. Among the comparisons, three trials reported a decline in eGFR with DPP-IV i, while one trial indicated that DPP-IV i prevented a decrease in the eGFR, and another trial found no significant difference between DPP-IV i and placebo. Additionally, the decline in UACR with DPP-IV i was not significantly different when compared to placebo (MD: -6.05, 95% CI [-14.63, 2.54], p = 0.17, I2 = 0%) [52, 55]. In Fig. 13C, the results of analyses were shown compared to controls; DPP-IV i did not significantly increase the risk of DR (RR: 0.78, 95% CI [0.41, 1.48], p = 0.44, I2 = 63%) [51, 56]. Figure 13D presents the analyses of DPP-IV i on the risk of DPN. Compared to controls, DPP-IV i significantly decreased the risk of DPN (RR: 0.10, 95% CI [0.08, 0.13], p < 0.0001) [51] Table 9.

Fig. 13.

Forest plot comparing the effects of DPP-IV inhibitors versus placebo on eGFR (A) with Funnel Plot (E); and UACR (B) with Funnel Plot (F); and versus non-DPP-IV inhibitors on the risk of diabetic retinopathy (C) with Funnel Plot (G) and diabetic peripheral neuropathy (D) in type 2 diabetes (T2D)

Table 9.

The study design and demographic characteristics of clinical trials of DPP-IVi

Author (Year)	Country	Study Design	Sample Size	Mean Age(years)	% Male	Intervention	Control/Comparator	Outcome Measure	Follow-up Duration (Wks)	References
Groop, P. H. (2014)	Multicenter	RCT	2143	57	50	Linagliptin	Placebo	eGFR; UACR	24	[52]
Laakso, M. (2015)	Multicenter	RCT	235	66	63	Linagliptin	Placebo	eGFR	52	[53]
Chung, Y. R. (2016)	Multicenter	RCT	82	56	52	DPP-IVi	Non-DPP-IVi	Fundus photographs	270	[56]
Engel, S. S. (2017)	Multicenter	RCT	3324	68	70	Sitagliptin	Placebo	eGFR; DR, DPN risk	118	[51]
Yoon, S. A. (2017)	Multicenter	RCT	148	62	58	gemigliptin	Placebo	eGFR; UACR	12	[55]
Perkovic, V. (2020)	Multicenter	RCT	6979	65	63	Linagliptin	Placebo	eGFR	119	[54]

Effects of GLP-1RA on DN, DPN, and DR

In Fig. 14, A and B, the results of analyses showed that GLP-1RA is associated with the decline of eGFR and UACR in DN. Compared with the placebo, GLP-1RA significantly prevented the decline of eGFR (MD: 0.62, 95% CI [0.46, 0.78], p < 0.00001, I2 = 0%) [57–59], (Fig. 14A). Regarding UACR, GLP-1RA did not significantly reduce it compared to the control (MD: -2.11, 95% CI [-6.95, 2.73], p = 0.39, I2 = 90%) [58, 59]. In Fig. 14C, the results of analyses showed the effects of GLP-1RA on the measures of DPN and CAN. Compared to insulin, GLP-1RA did not significantly alter the measures of CAN and DPN [60]. In Fig. 14D, the results of analyses showed that liraglutide, a GLP-1RA, did not significantly decrease the thickness of RNFL compared with the placebo (MD: -1.00, 95% CI [-4.37, 2.37], p = 0.56) [61]. Furthermore, the analyses indicated that semaglutide did not significantly increase the risk of DR in T2D compared to the placebo (RR: 1.20, 95% CI [0.99, 1.47], p = 0.07) [62] Table 10.

Fig. 14.

Forest plot comparing the effects of GLP-1RA versus placebo on eGFR with funnel plot (A, F), UACR (B, G); Exenatide versus insulin on measures of DPN (C); Liraglutide versus placebo on RNFL thickness of DR (D); and Semaglutide versus placebo on DR risk with funnel plot (E, H) in T2D

Table 10.

The study design and demographic characteristics of clinical trials of GLP-1RAs

Author (Year)	Country	Study Design	Sample Size	Mean Age(years)	% Male	Intervention	Control /Comparator	Outcome Measure	Follow-up Duration (Wks)	References
Gerstein, H. C. (2019)	Multicenter	RCT	9901	50	54	Dulaglutide 1.5 mg	Placebo	eGFR;	24	[57]
Mann, J. F. E. (2017)	Multicenter	RCT	9340	64	64	Liraglutide	Placebo	eGFR; UACR	207	[58]
Tuttle, K. R. (2017)	Multicenter	Pooled analysis of RCT	6005	56	51	Dulaglutide 1.5 mg	Placebo or Active comparators	eGFR; UACR	26	[59]
Jaiswal, M. (2015)	US	RCT	56	54	56	Exenatide bid	Glargine insulin	CAN	78	[60]
Nielsen, T. A. (2022)	Multicenter	RCT	48	49	79	Liraglutide Up to tolerated 1.8 mg	Placebo	Retinal RNFL Thickness	26	[61]
Vilsboll, T. (2018)	Multicenter	Pooled Analysis of RCT	SUSTAIN 1–5: 3899 SUSTAIN6: 3297	64	60	semaglutide (0.5;1.0 mg) once weekly	SUSTAIN 1–5: Placebo or active comparator; SUSTAIN 6: volume-matched placebo	Fundoscopy/fundus photography	SUSTAIN 1–5: 30; SUSTAIN 6: 104	[62]

Discussion

Our analysis indicates that there are numerous inherent distinctions in the ability of non-insulin anti-diabetic medications (ADMs) to reduce diabetic microvascular complications, as evidenced by various measurable indicators in DN, DPN, and DR, in addition to the effects of intensive glucose lowering. Our findings also suggest that specific targets, mechanisms, or pharmacological properties may relate to the effects of novel ADMs such as SGLT-2i, DPP-IVIs, and GLP-1RA, which provide additional benefits beyond hypoglycemia in reducing diabetic microvascular complications and require further exploration through clinical and basic studies. However, given the limited evidence available, the abilities and effects of traditional ADMs, such as MET, SUs, Repa, α-GIs, and TZDs, may necessitate more extensive sample inclusion and data collection in real-world practices. On the other hand, due to the different pathophysiological mechanisms and promoters involved in each microvascular complication, various ADMs may demonstrate different abilities and therapeutic properties in reducing them, which necessitates broadening their specific therapeutic fields and deepening investigations.

The effects of MET on the microvascular complications

Primarily, the analytical effects indicated that MET, as the first-line choice for T2D, did not significantly improve the eGFR compared to other drugs [12]. This may be due to insufficient data on monotherapy and the fact that multiple clinical trials assessed MET alone or in combination with other drugs as the comparative groups. In this trial, the MET group was compared with glybenclamide, a SUs drug, which had comparable baseline values. However, urinary microalbumin levels significantly improved (showed improvement in UAE, as did blood pressure, over 12 weeks in the MET group diagnosed with incipient nephropathy when compared to the baseline. In this study, the comparative drug MET did not demonstrate a superior ability to reduce UAE compared to SUs. Thus, the analyses did not confirm that MET could worsen renal function. It also assessed the efficacy of MET in preventing DPN and found that MET did not significantly affect CAN and may increase VDT [13].

The effects of SUs on the microvascular complications

Second, we found that SUs did not significantly deteriorate renal function, as indicated by a change in eGFR, and did not significantly improve UACR compared with the MET group [14]. However, SU improved UAE, when combined with MET and losartan, exhibited a substantial reduction in UAE compared to MET or losartan alone [15]. Regarding DR prevention, the analytical results indicated that SUs did not reduce the risk of DR, possibly due to the heterogeneity of the references (I2 = 65%) [16, 17]. This conclusion requires more robust evidence with less bias to establish that SUs may reduce the risk of DR.

The effects of Repa on the microvascular complications

In the next step, we tested the effects of Repa, a short-acting insulin secretagogue, on the UAE, revealing significant protective effects on DN comparable to the SUs in the same study [15].

The effects of α-GIs on the microvascular complications

Subsequently, the analysis indicated that α-GIs could significantly lower the UACR due to the potential bias present in this study and the lack of other evidence, which may only weakly demonstrate the DN-preventing property of α-GIs [18].

The effects of TZDs on the microvascular complications

Moreover, the TZDs showed no significant deterioration of eGFR due to the heterogeneity of the clinical trials (I² = 88%) [14, 19–22]. However, regardless of this heterogeneity, it consistently revealed that TZDs could significantly lower UACR compared to placebo in T2D (Fig. 8D) [23]. Additionally, TZDs could decrease UAE by two clinical trials proved [24, 25], although the reference indicated considerable heterogeneity (I² = 85%) and concerns about bias in these studies. The alleviation of DPN by TZDs also demonstrated improvements in VDT [26]. Regarding DR, the results indicated that TZDs did not significantly increase the risk of DR (I² = 0%) [27].; this may be particularly important in clinics, as rapid reductions in glycemia or HbA1c could exacerbate the symptoms of DR, yet TZDs do not significantly lower glycemia quickly, thus presenting a lesser association with the adverse effects of DR, and other mechanisms may require further investigation.

The effects of SGLT-2is on the microvascular complications

Furthermore, we analyzed the effects of SGLT-2i on diabetic microvascular prevention. The results showed that SGLT-2is had a protective effect on eGFR and UACR in T2D (Figs. 10 and 11) [28–49]. Although dapagliflozin did not demonstrate a substantial decrease in eGFR (Data not shown) [36, 43], other drugs consistently displayed renal protection, evidenced by improvements in eGFR. Regarding UACR, the studies consistently indicated that dapagliflozin could significantly reduce UACR (Data not shown) by -23.83 mg/g Cr [36, 43]. There were completed trials testing the effects of empagliflozin on UACR [44–47]; a final comparison between these two drugs may be prospective. The same situation applied to ertugliflozin, which was basically examined in only one trial of VERTIS in the current review [48, 49]. We found only evidence regarding canagliflozin’s protection against DPN, showing that it did not significantly increase the risk of DPN [50]; caution should be taken since canagliflozin should not be used in patients at risk for diabetic foot. Data from the CANVAS trial indicated an increased risk of lower-limb amputations (HR 1.97, 95% CI 1.41–2.75), particularly in patients with prior amputations or peripheral vascular disease, conditions often seen in diabetic foot cases; however, the CREDENCE trial suggested there was no difference in the risk of lower limb amputation [30, 63–65]. A recent subsequent US nationwide database analysis employing propensity score methods compared Canagliflozin vs. GLP-1RA and showed this trend was particularly evident in patients older than 65 with baseline cardiovascular diseases [66, 67]. Consequently, canagliflozin may not be suitable for this subgroup, despite its benefits for other microvascular complications like nephropathy [68]. In contrast, evidence from other SGLT-2is may be contradictory, warranting further investigation [69]. The pathophysiological mechanism underlying this complication has been referred to as the off-target effects of SGLT-2i classes, while some evidence suggested that it is more likely to occur in elderly patients with peripheral arterial disease (PAD), accounting for predisposed higher incidents of infection but not ischemia induced by SGLT-2i inhibition [70]. The renal and other protective activities against microvascular complications by SGLT-2i may involve multiple mechanisms such as tubule glomerular feedback, improvements in metabolic substrates, ketogenesis, autophagic pathway regulation, and modulation of autonomic nervous system activity, etc. Emerging trials should focus on alternative approaches beyond renal microvascular complications.

The effects of DPP-IVis on the microvascular complications

Finally, we analyzed two classes of drugs that target the same incretin pathway: DPP-IV inhibitors and GLP-1 receptor agonists. Although these two drugs address the same target, regardless of whether they elevate GLP-1 levels pharmacologically or physiologically, their effects exhibited multiple discrepancies. Most DPP-IV inhibitors, with the exception of linagliptin, are restricted by advanced renal insufficiency. Our analyses indicated that DPP-IV inhibitors did not decrease the eGFR (I² = 99%) and had limited effectiveness in reducing urinary albumin as measured by UACR (I² = 0) [51–55]. Furthermore, DPP-IV inhibitors did not significantly increase the risk of diabetic retinopathy (DR) [51, 56], although the heterogeneity in this comparison was high (I² = 63%). In contrast, limited evidence suggested that DPP-IV inhibitors may benefit the risk of diabetic peripheral neuropathy (DPN) [51]. Overall, these analyses demonstrated that DPP-IV inhibitors have therapeutic safety for type 2 diabetes (T2D).

The effects of GLP-1RA on the microvascular complications

Nonetheless, in GLP-1RA, we found that GLP-1RA could improve the eGFR compared to placebo (I2 = 0%) [57–59], while the trials with high heterogeneity (I2 = 90%) indicated it did not significantly affect UACR [58, 59]. This may be due to the selective criteria for a specific population, the quality, and the sample size in some included studies. More trials in the future are warranted to clarify the outcome of GLP-1RA on DN protection. We found that GLP-1RA did not significantly alter the measures of DPN and CAN [60], which may not confirm its therapeutic value for DPN protection due to the study sample and inherent bias. In DR protection, new trials reported that the weekly formulation of semaglutide did not significantly increase the risk of DR (I2 = 0%) in a series of SUSTAIN trials, while the only available data indicated that another long-acting agent [62], liraglutide, did not decrease the thickness of RNFL [61].

Strengths and limitations

We utilized extensive MESH terms combined with entry terms while searching major clinical databases and conducted a comprehensive screening of the results. We also tested our search content by individually searching one class of drugs with certain complications under the rules of Boolean Logic and found no significant omissions. There were some limitations to our study; first, it lacks evidence on other non-conventional drugs such as bromocriptine and pramlintide. These drugs have not been permitted for use in T2D in China; therefore, we did not have sufficient experience with them. Second, while we confirmed the effects of MET and SUs, there was insufficient evidence to compare them with no treatment alone for patients requiring intensive glycemic control, as most studies used SUs or MET as the control group. Consequently, we could not obtain mono-therapeutic effects of MET or SUs on the prevention of complications. We did not include the results of ongoing trials. New analytic methods, such as network meta-analysis, could be employed to simulate head-to-head comparisons of classes of drugs when necessary. In future analyses, we will update our techniques and incorporate new evidence.

Quality of the evidence

Although some early evidence of conventional anti-diabetic medications has limitations that may introduce bias, such as blinding and allocation, most of our evidence comes from RCTs with higher standards and clear descriptions of quality and bias control. We included references from high-impact journals and highly cited publications. We excluded secondary analytic articles such as reviews or meta-analyses, along with other types of publications, including case reports, conference abstracts, animal studies, and in vitro studies. We confirmed our analyses, although some effects with high heterogeneity could support our conclusions.

Potential biases in the review process

We made every effort to reduce the bias involved in the review process. However, potential biases may exist; for example, we generally included references with high impact and excluded papers without impact, such as grey literature. Additionally, due to language constraints, our search and discussion were primarily limited to English-language publications.

Agreements and disagreements with other studies or reviews

Our analysis is the first to summarize the effects of discrete anti-diabetic medications on diabetic microvascular complications. It has been reported that intensive glycemic control has only minimal impact on macrovascular diseases, although macrovascular disease is the leading cause of death among diabetic patients. However, unlike macrovascular diseases, microvascular diseases can be prevented by hypoglycemic therapy. Consequently, we adopted continuous indices to detect this effect, with fewer dichotomous outcomes such as mortality or survival rate.

Conclusions

Implications for practice

Our analysis revealed that various ADMs, particularly novel ones like SGLT-2 inhibitors and incretin-based therapies—including GLP-1 receptor agonists (GLP-1RA) and DPP-IV inhibitors—may benefit microvascular complications and expand upon traditional perspectives regarding intensive hypoglycemic actions. Future clinical practices could benefit from improved strategies that focus on specific pathophysiological links in T2DM. Additionally, this implies that in the early stages of T2DM, the advantages of novel ADMs or their combination in treatment may significantly enhance or delay the onset and progression of complications.

Implications for research

Given the limited evidence regarding research on diabetic microvascular complications, the unique mechanisms for each complication should be further explored, along with their links to effective therapeutic targets. Additionally, more advanced measures for DPN, DR, and other microvascular complications should be implemented to determine their extent and provide precise targets for the therapeutic value of various ADMs.

Author contributions

All authors (S. W.; Y. Y; YY. L; CL.X; LJ. C; YS R; YJ. H; XC. L; M. G; XL. Y; DX. X; CX. W; LG. Z) contributed to the reported findings, including study conception, study design, execution, data acquisition, analysis, and interpretation, and drafting, revising, or critically reviewing the article. All authors approved the final version of the manuscript to be published.

Funding

This work was supported by the Fudan Zhangjiang Clinical Medicine Innovation Fund Project (KP0202118), Integrated Traditional Chinese and Western Medicine (YC-2023-0404), Fudan Good Practice Program of Teaching and Learning (FD2023A227), Talents Training Program of Shanghai Pudong Hospital (YQ202101); Project of Key Medical Discipline of Pudong Hospital of Fudan University (Zdxk2020-11), Project of Key Medical Specialty and Treatment Center of Pudong Hospital of Fudan University (Zdzk2020-24), Integrative Medicine special fund of Shanghai Municipal Health Planning Committee (ZHYY- ZXYJHZX-2-201712), Special Department Fund of the Pudong New Area Health Planning Commission (PWZzk2017-03), Outstanding Leaders Training Program of Pudong Health Bureau of Shanghai (PWR12014-06), Pudong New Area Clinical Plateau Discipline Project (PWYgy-2021-03), the Natural Science Foundation of China (21675034), National Natural Science Foundation of China (81370932), Shanghai Natural Science Foundation (19ZR1447500), Pudong New Area Clinical Characteristic Discipline Project (PWYts2021-11), Pudong New Area Clinical Characteristic Discipline Project (PWYts2021-01), Wenzhou Medical University Education Grant (JG2021197).

Data availability

All data generated or analyzed during this study are included in this article.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.